2 * Copyright (C) 2010-2011 Neil Brown
3 * Copyright (C) 2010-2016 Red Hat, Inc. All rights reserved.
5 * This file is released under the GPL.
8 #include <linux/slab.h>
9 #include <linux/module.h>
17 #include <linux/device-mapper.h>
19 #define DM_MSG_PREFIX "raid"
20 #define MAX_RAID_DEVICES 253 /* md-raid kernel limit */
23 * Minimum sectors of free reshape space per raid device
25 #define MIN_FREE_RESHAPE_SPACE to_sector(4*4096)
27 static bool devices_handle_discard_safely
= false;
30 * The following flags are used by dm-raid.c to set up the array state.
31 * They must be cleared before md_run is called.
33 #define FirstUse 10 /* rdev flag */
37 * Two DM devices, one to hold metadata and one to hold the
38 * actual data/parity. The reason for this is to not confuse
39 * ti->len and give more flexibility in altering size and
42 * While it is possible for this device to be associated
43 * with a different physical device than the data_dev, it
44 * is intended for it to be the same.
45 * |--------- Physical Device ---------|
46 * |- meta_dev -|------ data_dev ------|
48 struct dm_dev
*meta_dev
;
49 struct dm_dev
*data_dev
;
54 * Bits for establishing rs->ctr_flags
59 #define __CTR_FLAG_SYNC 0 /* 1 */ /* Not with raid0! */
60 #define __CTR_FLAG_NOSYNC 1 /* 1 */ /* Not with raid0! */
61 #define __CTR_FLAG_REBUILD 2 /* 2 */ /* Not with raid0! */
62 #define __CTR_FLAG_DAEMON_SLEEP 3 /* 2 */ /* Not with raid0! */
63 #define __CTR_FLAG_MIN_RECOVERY_RATE 4 /* 2 */ /* Not with raid0! */
64 #define __CTR_FLAG_MAX_RECOVERY_RATE 5 /* 2 */ /* Not with raid0! */
65 #define __CTR_FLAG_MAX_WRITE_BEHIND 6 /* 2 */ /* Only with raid1! */
66 #define __CTR_FLAG_WRITE_MOSTLY 7 /* 2 */ /* Only with raid1! */
67 #define __CTR_FLAG_STRIPE_CACHE 8 /* 2 */ /* Only with raid4/5/6! */
68 #define __CTR_FLAG_REGION_SIZE 9 /* 2 */ /* Not with raid0! */
69 #define __CTR_FLAG_RAID10_COPIES 10 /* 2 */ /* Only with raid10 */
70 #define __CTR_FLAG_RAID10_FORMAT 11 /* 2 */ /* Only with raid10 */
72 #define __CTR_FLAG_DELTA_DISKS 12 /* 2 */ /* Only with reshapable raid1/4/5/6/10! */
73 #define __CTR_FLAG_DATA_OFFSET 13 /* 2 */ /* Only with reshapable raid4/5/6/10! */
74 #define __CTR_FLAG_RAID10_USE_NEAR_SETS 14 /* 2 */ /* Only with raid10! */
77 * Flags for rs->ctr_flags field.
79 #define CTR_FLAG_SYNC (1 << __CTR_FLAG_SYNC)
80 #define CTR_FLAG_NOSYNC (1 << __CTR_FLAG_NOSYNC)
81 #define CTR_FLAG_REBUILD (1 << __CTR_FLAG_REBUILD)
82 #define CTR_FLAG_DAEMON_SLEEP (1 << __CTR_FLAG_DAEMON_SLEEP)
83 #define CTR_FLAG_MIN_RECOVERY_RATE (1 << __CTR_FLAG_MIN_RECOVERY_RATE)
84 #define CTR_FLAG_MAX_RECOVERY_RATE (1 << __CTR_FLAG_MAX_RECOVERY_RATE)
85 #define CTR_FLAG_MAX_WRITE_BEHIND (1 << __CTR_FLAG_MAX_WRITE_BEHIND)
86 #define CTR_FLAG_WRITE_MOSTLY (1 << __CTR_FLAG_WRITE_MOSTLY)
87 #define CTR_FLAG_STRIPE_CACHE (1 << __CTR_FLAG_STRIPE_CACHE)
88 #define CTR_FLAG_REGION_SIZE (1 << __CTR_FLAG_REGION_SIZE)
89 #define CTR_FLAG_RAID10_COPIES (1 << __CTR_FLAG_RAID10_COPIES)
90 #define CTR_FLAG_RAID10_FORMAT (1 << __CTR_FLAG_RAID10_FORMAT)
91 #define CTR_FLAG_DELTA_DISKS (1 << __CTR_FLAG_DELTA_DISKS)
92 #define CTR_FLAG_DATA_OFFSET (1 << __CTR_FLAG_DATA_OFFSET)
93 #define CTR_FLAG_RAID10_USE_NEAR_SETS (1 << __CTR_FLAG_RAID10_USE_NEAR_SETS)
96 * Definitions of various constructor flags to
97 * be used in checks of valid / invalid flags
100 /* Define all any sync flags */
101 #define CTR_FLAGS_ANY_SYNC (CTR_FLAG_SYNC | CTR_FLAG_NOSYNC)
103 /* Define flags for options without argument (e.g. 'nosync') */
104 #define CTR_FLAG_OPTIONS_NO_ARGS (CTR_FLAGS_ANY_SYNC | \
105 CTR_FLAG_RAID10_USE_NEAR_SETS)
107 /* Define flags for options with one argument (e.g. 'delta_disks +2') */
108 #define CTR_FLAG_OPTIONS_ONE_ARG (CTR_FLAG_REBUILD | \
109 CTR_FLAG_WRITE_MOSTLY | \
110 CTR_FLAG_DAEMON_SLEEP | \
111 CTR_FLAG_MIN_RECOVERY_RATE | \
112 CTR_FLAG_MAX_RECOVERY_RATE | \
113 CTR_FLAG_MAX_WRITE_BEHIND | \
114 CTR_FLAG_STRIPE_CACHE | \
115 CTR_FLAG_REGION_SIZE | \
116 CTR_FLAG_RAID10_COPIES | \
117 CTR_FLAG_RAID10_FORMAT | \
118 CTR_FLAG_DELTA_DISKS | \
119 CTR_FLAG_DATA_OFFSET)
121 /* Valid options definitions per raid level... */
123 /* "raid0" does only accept data offset */
124 #define RAID0_VALID_FLAGS (CTR_FLAG_DATA_OFFSET)
126 /* "raid1" does not accept stripe cache, data offset, delta_disks or any raid10 options */
127 #define RAID1_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \
129 CTR_FLAG_WRITE_MOSTLY | \
130 CTR_FLAG_DAEMON_SLEEP | \
131 CTR_FLAG_MIN_RECOVERY_RATE | \
132 CTR_FLAG_MAX_RECOVERY_RATE | \
133 CTR_FLAG_MAX_WRITE_BEHIND | \
134 CTR_FLAG_REGION_SIZE | \
135 CTR_FLAG_DELTA_DISKS | \
136 CTR_FLAG_DATA_OFFSET)
138 /* "raid10" does not accept any raid1 or stripe cache options */
139 #define RAID10_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \
141 CTR_FLAG_DAEMON_SLEEP | \
142 CTR_FLAG_MIN_RECOVERY_RATE | \
143 CTR_FLAG_MAX_RECOVERY_RATE | \
144 CTR_FLAG_REGION_SIZE | \
145 CTR_FLAG_RAID10_COPIES | \
146 CTR_FLAG_RAID10_FORMAT | \
147 CTR_FLAG_DELTA_DISKS | \
148 CTR_FLAG_DATA_OFFSET | \
149 CTR_FLAG_RAID10_USE_NEAR_SETS)
152 * "raid4/5/6" do not accept any raid1 or raid10 specific options
154 * "raid6" does not accept "nosync", because it is not guaranteed
155 * that both parity and q-syndrome are being written properly with
158 #define RAID45_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \
160 CTR_FLAG_DAEMON_SLEEP | \
161 CTR_FLAG_MIN_RECOVERY_RATE | \
162 CTR_FLAG_MAX_RECOVERY_RATE | \
163 CTR_FLAG_MAX_WRITE_BEHIND | \
164 CTR_FLAG_STRIPE_CACHE | \
165 CTR_FLAG_REGION_SIZE | \
166 CTR_FLAG_DELTA_DISKS | \
167 CTR_FLAG_DATA_OFFSET)
169 #define RAID6_VALID_FLAGS (CTR_FLAG_SYNC | \
171 CTR_FLAG_DAEMON_SLEEP | \
172 CTR_FLAG_MIN_RECOVERY_RATE | \
173 CTR_FLAG_MAX_RECOVERY_RATE | \
174 CTR_FLAG_MAX_WRITE_BEHIND | \
175 CTR_FLAG_STRIPE_CACHE | \
176 CTR_FLAG_REGION_SIZE | \
177 CTR_FLAG_DELTA_DISKS | \
178 CTR_FLAG_DATA_OFFSET)
179 /* ...valid options definitions per raid level */
182 * Flags for rs->runtime_flags field
183 * (RT_FLAG prefix meaning "runtime flag")
185 * These are all internal and used to define runtime state,
186 * e.g. to prevent another resume from preresume processing
187 * the raid set all over again.
189 #define RT_FLAG_RS_PRERESUMED 0
190 #define RT_FLAG_RS_RESUMED 1
191 #define RT_FLAG_RS_BITMAP_LOADED 2
192 #define RT_FLAG_UPDATE_SBS 3
193 #define RT_FLAG_RESHAPE_RS 4
195 /* Array elements of 64 bit needed for rebuild/failed disk bits */
196 #define DISKS_ARRAY_ELEMS ((MAX_RAID_DEVICES + (sizeof(uint64_t) * 8 - 1)) / sizeof(uint64_t) / 8)
199 * raid set level, layout and chunk sectors backup/restore
204 int new_chunk_sectors
;
208 struct dm_target
*ti
;
210 uint32_t bitmap_loaded
;
211 uint32_t stripe_cache_entries
;
212 unsigned long ctr_flags
;
213 unsigned long runtime_flags
;
215 uint64_t rebuild_disks
[DISKS_ARRAY_ELEMS
];
221 int requested_bitmap_chunk_sectors
;
224 struct raid_type
*raid_type
;
225 struct dm_target_callbacks callbacks
;
227 struct raid_dev dev
[0];
230 static void rs_config_backup(struct raid_set
*rs
, struct rs_layout
*l
)
232 struct mddev
*mddev
= &rs
->md
;
234 l
->new_level
= mddev
->new_level
;
235 l
->new_layout
= mddev
->new_layout
;
236 l
->new_chunk_sectors
= mddev
->new_chunk_sectors
;
239 static void rs_config_restore(struct raid_set
*rs
, struct rs_layout
*l
)
241 struct mddev
*mddev
= &rs
->md
;
243 mddev
->new_level
= l
->new_level
;
244 mddev
->new_layout
= l
->new_layout
;
245 mddev
->new_chunk_sectors
= l
->new_chunk_sectors
;
248 /* raid10 algorithms (i.e. formats) */
249 #define ALGORITHM_RAID10_DEFAULT 0
250 #define ALGORITHM_RAID10_NEAR 1
251 #define ALGORITHM_RAID10_OFFSET 2
252 #define ALGORITHM_RAID10_FAR 3
254 /* Supported raid types and properties. */
255 static struct raid_type
{
256 const char *name
; /* RAID algorithm. */
257 const char *descr
; /* Descriptor text for logging. */
258 const unsigned int parity_devs
; /* # of parity devices. */
259 const unsigned int minimal_devs
;/* minimal # of devices in set. */
260 const unsigned int level
; /* RAID level. */
261 const unsigned int algorithm
; /* RAID algorithm. */
263 {"raid0", "raid0 (striping)", 0, 2, 0, 0 /* NONE */},
264 {"raid1", "raid1 (mirroring)", 0, 2, 1, 0 /* NONE */},
265 {"raid10_far", "raid10 far (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_FAR
},
266 {"raid10_offset", "raid10 offset (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_OFFSET
},
267 {"raid10_near", "raid10 near (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_NEAR
},
268 {"raid10", "raid10 (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_DEFAULT
},
269 {"raid4", "raid4 (dedicated last parity disk)", 1, 2, 4, ALGORITHM_PARITY_N
}, /* raid4 layout = raid5_n */
270 {"raid5_n", "raid5 (dedicated last parity disk)", 1, 2, 5, ALGORITHM_PARITY_N
},
271 {"raid5_ls", "raid5 (left symmetric)", 1, 2, 5, ALGORITHM_LEFT_SYMMETRIC
},
272 {"raid5_rs", "raid5 (right symmetric)", 1, 2, 5, ALGORITHM_RIGHT_SYMMETRIC
},
273 {"raid5_la", "raid5 (left asymmetric)", 1, 2, 5, ALGORITHM_LEFT_ASYMMETRIC
},
274 {"raid5_ra", "raid5 (right asymmetric)", 1, 2, 5, ALGORITHM_RIGHT_ASYMMETRIC
},
275 {"raid6_zr", "raid6 (zero restart)", 2, 4, 6, ALGORITHM_ROTATING_ZERO_RESTART
},
276 {"raid6_nr", "raid6 (N restart)", 2, 4, 6, ALGORITHM_ROTATING_N_RESTART
},
277 {"raid6_nc", "raid6 (N continue)", 2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE
},
278 {"raid6_n_6", "raid6 (dedicated parity/Q n/6)", 2, 4, 6, ALGORITHM_PARITY_N_6
},
279 {"raid6_ls_6", "raid6 (left symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_SYMMETRIC_6
},
280 {"raid6_rs_6", "raid6 (right symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_SYMMETRIC_6
},
281 {"raid6_la_6", "raid6 (left asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_ASYMMETRIC_6
},
282 {"raid6_ra_6", "raid6 (right asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_ASYMMETRIC_6
}
285 /* True, if @v is in inclusive range [@min, @max] */
286 static bool __within_range(long v
, long min
, long max
)
288 return v
>= min
&& v
<= max
;
291 /* All table line arguments are defined here */
292 static struct arg_name_flag
{
293 const unsigned long flag
;
295 } __arg_name_flags
[] = {
296 { CTR_FLAG_SYNC
, "sync"},
297 { CTR_FLAG_NOSYNC
, "nosync"},
298 { CTR_FLAG_REBUILD
, "rebuild"},
299 { CTR_FLAG_DAEMON_SLEEP
, "daemon_sleep"},
300 { CTR_FLAG_MIN_RECOVERY_RATE
, "min_recovery_rate"},
301 { CTR_FLAG_MAX_RECOVERY_RATE
, "max_recovery_rate"},
302 { CTR_FLAG_MAX_WRITE_BEHIND
, "max_write_behind"},
303 { CTR_FLAG_WRITE_MOSTLY
, "write_mostly"},
304 { CTR_FLAG_STRIPE_CACHE
, "stripe_cache"},
305 { CTR_FLAG_REGION_SIZE
, "region_size"},
306 { CTR_FLAG_RAID10_COPIES
, "raid10_copies"},
307 { CTR_FLAG_RAID10_FORMAT
, "raid10_format"},
308 { CTR_FLAG_DATA_OFFSET
, "data_offset"},
309 { CTR_FLAG_DELTA_DISKS
, "delta_disks"},
310 { CTR_FLAG_RAID10_USE_NEAR_SETS
, "raid10_use_near_sets"},
313 /* Return argument name string for given @flag */
314 static const char *dm_raid_arg_name_by_flag(const uint32_t flag
)
316 if (hweight32(flag
) == 1) {
317 struct arg_name_flag
*anf
= __arg_name_flags
+ ARRAY_SIZE(__arg_name_flags
);
319 while (anf
-- > __arg_name_flags
)
320 if (flag
& anf
->flag
)
324 DMERR("%s called with more than one flag!", __func__
);
330 * Bool helpers to test for various raid levels of a raid set.
331 * It's level as reported by the superblock rather than
332 * the requested raid_type passed to the constructor.
334 /* Return true, if raid set in @rs is raid0 */
335 static bool rs_is_raid0(struct raid_set
*rs
)
337 return !rs
->md
.level
;
340 /* Return true, if raid set in @rs is raid1 */
341 static bool rs_is_raid1(struct raid_set
*rs
)
343 return rs
->md
.level
== 1;
346 /* Return true, if raid set in @rs is raid10 */
347 static bool rs_is_raid10(struct raid_set
*rs
)
349 return rs
->md
.level
== 10;
352 /* Return true, if raid set in @rs is level 6 */
353 static bool rs_is_raid6(struct raid_set
*rs
)
355 return rs
->md
.level
== 6;
358 /* Return true, if raid set in @rs is level 4, 5 or 6 */
359 static bool rs_is_raid456(struct raid_set
*rs
)
361 return __within_range(rs
->md
.level
, 4, 6);
364 /* Return true, if raid set in @rs is reshapable */
365 static bool __is_raid10_far(int layout
);
366 static bool rs_is_reshapable(struct raid_set
*rs
)
368 return rs_is_raid456(rs
) ||
369 (rs_is_raid10(rs
) && !__is_raid10_far(rs
->md
.new_layout
));
372 /* Return true, if raid set in @rs is recovering */
373 static bool rs_is_recovering(struct raid_set
*rs
)
375 return rs
->md
.recovery_cp
< rs
->dev
[0].rdev
.sectors
;
378 /* Return true, if raid set in @rs is reshaping */
379 static bool rs_is_reshaping(struct raid_set
*rs
)
381 return rs
->md
.reshape_position
!= MaxSector
;
385 * bool helpers to test for various raid levels of a raid type @rt
388 /* Return true, if raid type in @rt is raid0 */
389 static bool rt_is_raid0(struct raid_type
*rt
)
394 /* Return true, if raid type in @rt is raid1 */
395 static bool rt_is_raid1(struct raid_type
*rt
)
397 return rt
->level
== 1;
400 /* Return true, if raid type in @rt is raid10 */
401 static bool rt_is_raid10(struct raid_type
*rt
)
403 return rt
->level
== 10;
406 /* Return true, if raid type in @rt is raid4/5 */
407 static bool rt_is_raid45(struct raid_type
*rt
)
409 return __within_range(rt
->level
, 4, 5);
412 /* Return true, if raid type in @rt is raid6 */
413 static bool rt_is_raid6(struct raid_type
*rt
)
415 return rt
->level
== 6;
418 /* Return true, if raid type in @rt is raid4/5/6 */
419 static bool rt_is_raid456(struct raid_type
*rt
)
421 return __within_range(rt
->level
, 4, 6);
423 /* END: raid level bools */
425 /* Return valid ctr flags for the raid level of @rs */
426 static unsigned long __valid_flags(struct raid_set
*rs
)
428 if (rt_is_raid0(rs
->raid_type
))
429 return RAID0_VALID_FLAGS
;
430 else if (rt_is_raid1(rs
->raid_type
))
431 return RAID1_VALID_FLAGS
;
432 else if (rt_is_raid10(rs
->raid_type
))
433 return RAID10_VALID_FLAGS
;
434 else if (rt_is_raid45(rs
->raid_type
))
435 return RAID45_VALID_FLAGS
;
436 else if (rt_is_raid6(rs
->raid_type
))
437 return RAID6_VALID_FLAGS
;
443 * Check for valid flags set on @rs
445 * Has to be called after parsing of the ctr flags!
447 static int rs_check_for_valid_flags(struct raid_set
*rs
)
449 if (rs
->ctr_flags
& ~__valid_flags(rs
)) {
450 rs
->ti
->error
= "Invalid flags combination";
457 /* MD raid10 bit definitions and helpers */
458 #define RAID10_OFFSET (1 << 16) /* stripes with data copies area adjacent on devices */
459 #define RAID10_BROCKEN_USE_FAR_SETS (1 << 17) /* Broken in raid10.c: use sets instead of whole stripe rotation */
460 #define RAID10_USE_FAR_SETS (1 << 18) /* Use sets instead of whole stripe rotation */
461 #define RAID10_FAR_COPIES_SHIFT 8 /* raid10 # far copies shift (2nd byte of layout) */
463 /* Return md raid10 near copies for @layout */
464 static unsigned int __raid10_near_copies(int layout
)
466 return layout
& 0xFF;
469 /* Return md raid10 far copies for @layout */
470 static unsigned int __raid10_far_copies(int layout
)
472 return __raid10_near_copies(layout
>> RAID10_FAR_COPIES_SHIFT
);
475 /* Return true if md raid10 offset for @layout */
476 static bool __is_raid10_offset(int layout
)
478 return !!(layout
& RAID10_OFFSET
);
481 /* Return true if md raid10 near for @layout */
482 static bool __is_raid10_near(int layout
)
484 return !__is_raid10_offset(layout
) && __raid10_near_copies(layout
) > 1;
487 /* Return true if md raid10 far for @layout */
488 static bool __is_raid10_far(int layout
)
490 return !__is_raid10_offset(layout
) && __raid10_far_copies(layout
) > 1;
493 /* Return md raid10 layout string for @layout */
494 static const char *raid10_md_layout_to_format(int layout
)
497 * Bit 16 stands for "offset"
498 * (i.e. adjacent stripes hold copies)
500 * Refer to MD's raid10.c for details
502 if (__is_raid10_offset(layout
))
505 if (__raid10_near_copies(layout
) > 1)
508 WARN_ON(__raid10_far_copies(layout
) < 2);
513 /* Return md raid10 algorithm for @name */
514 static int raid10_name_to_format(const char *name
)
516 if (!strcasecmp(name
, "near"))
517 return ALGORITHM_RAID10_NEAR
;
518 else if (!strcasecmp(name
, "offset"))
519 return ALGORITHM_RAID10_OFFSET
;
520 else if (!strcasecmp(name
, "far"))
521 return ALGORITHM_RAID10_FAR
;
526 /* Return md raid10 copies for @layout */
527 static unsigned int raid10_md_layout_to_copies(int layout
)
529 return max(__raid10_near_copies(layout
), __raid10_far_copies(layout
));
532 /* Return md raid10 format id for @format string */
533 static int raid10_format_to_md_layout(struct raid_set
*rs
,
534 unsigned int algorithm
,
537 unsigned int n
= 1, f
= 1, r
= 0;
540 * MD resilienece flaw:
542 * enabling use_far_sets for far/offset formats causes copies
543 * to be colocated on the same devs together with their origins!
545 * -> disable it for now in the definition above
547 if (algorithm
== ALGORITHM_RAID10_DEFAULT
||
548 algorithm
== ALGORITHM_RAID10_NEAR
)
551 else if (algorithm
== ALGORITHM_RAID10_OFFSET
) {
554 if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS
, &rs
->ctr_flags
))
555 r
|= RAID10_USE_FAR_SETS
;
557 } else if (algorithm
== ALGORITHM_RAID10_FAR
) {
560 if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS
, &rs
->ctr_flags
))
561 r
|= RAID10_USE_FAR_SETS
;
566 return r
| (f
<< RAID10_FAR_COPIES_SHIFT
) | n
;
568 /* END: MD raid10 bit definitions and helpers */
570 /* Check for any of the raid10 algorithms */
571 static bool __got_raid10(struct raid_type
*rtp
, const int layout
)
573 if (rtp
->level
== 10) {
574 switch (rtp
->algorithm
) {
575 case ALGORITHM_RAID10_DEFAULT
:
576 case ALGORITHM_RAID10_NEAR
:
577 return __is_raid10_near(layout
);
578 case ALGORITHM_RAID10_OFFSET
:
579 return __is_raid10_offset(layout
);
580 case ALGORITHM_RAID10_FAR
:
581 return __is_raid10_far(layout
);
590 /* Return raid_type for @name */
591 static struct raid_type
*get_raid_type(const char *name
)
593 struct raid_type
*rtp
= raid_types
+ ARRAY_SIZE(raid_types
);
595 while (rtp
-- > raid_types
)
596 if (!strcasecmp(rtp
->name
, name
))
602 /* Return raid_type for @name based derived from @level and @layout */
603 static struct raid_type
*get_raid_type_by_ll(const int level
, const int layout
)
605 struct raid_type
*rtp
= raid_types
+ ARRAY_SIZE(raid_types
);
607 while (rtp
-- > raid_types
) {
608 /* RAID10 special checks based on @layout flags/properties */
609 if (rtp
->level
== level
&&
610 (__got_raid10(rtp
, layout
) || rtp
->algorithm
== layout
))
618 * Conditionally change bdev capacity of @rs
619 * in case of a disk add/remove reshape
621 static void rs_set_capacity(struct raid_set
*rs
)
623 struct mddev
*mddev
= &rs
->md
;
624 struct md_rdev
*rdev
;
625 struct gendisk
*gendisk
= dm_disk(dm_table_get_md(rs
->ti
->table
));
628 * raid10 sets rdev->sector to the device size, which
629 * is unintended in case of out-of-place reshaping
631 rdev_for_each(rdev
, mddev
)
632 rdev
->sectors
= mddev
->dev_sectors
;
634 set_capacity(gendisk
, mddev
->array_sectors
);
635 revalidate_disk(gendisk
);
639 * Set the mddev properties in @rs to the current
640 * ones retrieved from the freshest superblock
642 static void rs_set_cur(struct raid_set
*rs
)
644 struct mddev
*mddev
= &rs
->md
;
646 mddev
->new_level
= mddev
->level
;
647 mddev
->new_layout
= mddev
->layout
;
648 mddev
->new_chunk_sectors
= mddev
->chunk_sectors
;
652 * Set the mddev properties in @rs to the new
653 * ones requested by the ctr
655 static void rs_set_new(struct raid_set
*rs
)
657 struct mddev
*mddev
= &rs
->md
;
659 mddev
->level
= mddev
->new_level
;
660 mddev
->layout
= mddev
->new_layout
;
661 mddev
->chunk_sectors
= mddev
->new_chunk_sectors
;
662 mddev
->raid_disks
= rs
->raid_disks
;
663 mddev
->delta_disks
= 0;
666 static struct raid_set
*raid_set_alloc(struct dm_target
*ti
, struct raid_type
*raid_type
,
667 unsigned int raid_devs
)
672 if (raid_devs
<= raid_type
->parity_devs
) {
673 ti
->error
= "Insufficient number of devices";
674 return ERR_PTR(-EINVAL
);
677 rs
= kzalloc(sizeof(*rs
) + raid_devs
* sizeof(rs
->dev
[0]), GFP_KERNEL
);
679 ti
->error
= "Cannot allocate raid context";
680 return ERR_PTR(-ENOMEM
);
685 rs
->raid_disks
= raid_devs
;
689 rs
->raid_type
= raid_type
;
690 rs
->stripe_cache_entries
= 256;
691 rs
->md
.raid_disks
= raid_devs
;
692 rs
->md
.level
= raid_type
->level
;
693 rs
->md
.new_level
= rs
->md
.level
;
694 rs
->md
.layout
= raid_type
->algorithm
;
695 rs
->md
.new_layout
= rs
->md
.layout
;
696 rs
->md
.delta_disks
= 0;
697 rs
->md
.recovery_cp
= MaxSector
;
699 for (i
= 0; i
< raid_devs
; i
++)
700 md_rdev_init(&rs
->dev
[i
].rdev
);
703 * Remaining items to be initialized by further RAID params:
706 * rs->md.chunk_sectors
707 * rs->md.new_chunk_sectors
714 static void raid_set_free(struct raid_set
*rs
)
718 for (i
= 0; i
< rs
->raid_disks
; i
++) {
719 if (rs
->dev
[i
].meta_dev
)
720 dm_put_device(rs
->ti
, rs
->dev
[i
].meta_dev
);
721 md_rdev_clear(&rs
->dev
[i
].rdev
);
722 if (rs
->dev
[i
].data_dev
)
723 dm_put_device(rs
->ti
, rs
->dev
[i
].data_dev
);
730 * For every device we have two words
731 * <meta_dev>: meta device name or '-' if missing
732 * <data_dev>: data device name or '-' if missing
734 * The following are permitted:
737 * <meta_dev> <data_dev>
739 * The following is not allowed:
742 * This code parses those words. If there is a failure,
743 * the caller must use raid_set_free() to unwind the operations.
745 static int parse_dev_params(struct raid_set
*rs
, struct dm_arg_set
*as
)
749 int metadata_available
= 0;
753 /* Put off the number of raid devices argument to get to dev pairs */
754 arg
= dm_shift_arg(as
);
758 for (i
= 0; i
< rs
->raid_disks
; i
++) {
759 rs
->dev
[i
].rdev
.raid_disk
= i
;
761 rs
->dev
[i
].meta_dev
= NULL
;
762 rs
->dev
[i
].data_dev
= NULL
;
765 * There are no offsets, since there is a separate device
766 * for data and metadata.
768 rs
->dev
[i
].rdev
.data_offset
= 0;
769 rs
->dev
[i
].rdev
.mddev
= &rs
->md
;
771 arg
= dm_shift_arg(as
);
775 if (strcmp(arg
, "-")) {
776 r
= dm_get_device(rs
->ti
, arg
, dm_table_get_mode(rs
->ti
->table
),
777 &rs
->dev
[i
].meta_dev
);
779 rs
->ti
->error
= "RAID metadata device lookup failure";
783 rs
->dev
[i
].rdev
.sb_page
= alloc_page(GFP_KERNEL
);
784 if (!rs
->dev
[i
].rdev
.sb_page
) {
785 rs
->ti
->error
= "Failed to allocate superblock page";
790 arg
= dm_shift_arg(as
);
794 if (!strcmp(arg
, "-")) {
795 if (!test_bit(In_sync
, &rs
->dev
[i
].rdev
.flags
) &&
796 (!rs
->dev
[i
].rdev
.recovery_offset
)) {
797 rs
->ti
->error
= "Drive designated for rebuild not specified";
801 if (rs
->dev
[i
].meta_dev
) {
802 rs
->ti
->error
= "No data device supplied with metadata device";
809 r
= dm_get_device(rs
->ti
, arg
, dm_table_get_mode(rs
->ti
->table
),
810 &rs
->dev
[i
].data_dev
);
812 rs
->ti
->error
= "RAID device lookup failure";
816 if (rs
->dev
[i
].meta_dev
) {
817 metadata_available
= 1;
818 rs
->dev
[i
].rdev
.meta_bdev
= rs
->dev
[i
].meta_dev
->bdev
;
820 rs
->dev
[i
].rdev
.bdev
= rs
->dev
[i
].data_dev
->bdev
;
821 list_add_tail(&rs
->dev
[i
].rdev
.same_set
, &rs
->md
.disks
);
822 if (!test_bit(In_sync
, &rs
->dev
[i
].rdev
.flags
))
826 if (metadata_available
) {
828 rs
->md
.persistent
= 1;
829 rs
->md
.major_version
= 2;
830 } else if (rebuild
&& !rs
->md
.recovery_cp
) {
832 * Without metadata, we will not be able to tell if the array
833 * is in-sync or not - we must assume it is not. Therefore,
834 * it is impossible to rebuild a drive.
836 * Even if there is metadata, the on-disk information may
837 * indicate that the array is not in-sync and it will then
840 * User could specify 'nosync' option if desperate.
842 rs
->ti
->error
= "Unable to rebuild drive while array is not in-sync";
850 * validate_region_size
852 * @region_size: region size in sectors. If 0, pick a size (4MiB default).
854 * Set rs->md.bitmap_info.chunksize (which really refers to 'region size').
855 * Ensure that (ti->len/region_size < 2^21) - required by MD bitmap.
857 * Returns: 0 on success, -EINVAL on failure.
859 static int validate_region_size(struct raid_set
*rs
, unsigned long region_size
)
861 unsigned long min_region_size
= rs
->ti
->len
/ (1 << 21);
865 * Choose a reasonable default. All figures in sectors.
867 if (min_region_size
> (1 << 13)) {
868 /* If not a power of 2, make it the next power of 2 */
869 region_size
= roundup_pow_of_two(min_region_size
);
870 DMINFO("Choosing default region size of %lu sectors",
873 DMINFO("Choosing default region size of 4MiB");
874 region_size
= 1 << 13; /* sectors */
878 * Validate user-supplied value.
880 if (region_size
> rs
->ti
->len
) {
881 rs
->ti
->error
= "Supplied region size is too large";
885 if (region_size
< min_region_size
) {
886 DMERR("Supplied region_size (%lu sectors) below minimum (%lu)",
887 region_size
, min_region_size
);
888 rs
->ti
->error
= "Supplied region size is too small";
892 if (!is_power_of_2(region_size
)) {
893 rs
->ti
->error
= "Region size is not a power of 2";
897 if (region_size
< rs
->md
.chunk_sectors
) {
898 rs
->ti
->error
= "Region size is smaller than the chunk size";
904 * Convert sectors to bytes.
906 rs
->md
.bitmap_info
.chunksize
= to_bytes(region_size
);
912 * validate_raid_redundancy
915 * Determine if there are enough devices in the array that haven't
916 * failed (or are being rebuilt) to form a usable array.
918 * Returns: 0 on success, -EINVAL on failure.
920 static int validate_raid_redundancy(struct raid_set
*rs
)
922 unsigned int i
, rebuild_cnt
= 0;
923 unsigned int rebuilds_per_group
= 0, copies
;
924 unsigned int group_size
, last_group_start
;
926 for (i
= 0; i
< rs
->md
.raid_disks
; i
++)
927 if (!test_bit(In_sync
, &rs
->dev
[i
].rdev
.flags
) ||
928 !rs
->dev
[i
].rdev
.sb_page
)
931 switch (rs
->raid_type
->level
) {
933 if (rebuild_cnt
>= rs
->md
.raid_disks
)
939 if (rebuild_cnt
> rs
->raid_type
->parity_devs
)
943 copies
= raid10_md_layout_to_copies(rs
->md
.new_layout
);
944 if (rebuild_cnt
< copies
)
948 * It is possible to have a higher rebuild count for RAID10,
949 * as long as the failed devices occur in different mirror
950 * groups (i.e. different stripes).
952 * When checking "near" format, make sure no adjacent devices
953 * have failed beyond what can be handled. In addition to the
954 * simple case where the number of devices is a multiple of the
955 * number of copies, we must also handle cases where the number
956 * of devices is not a multiple of the number of copies.
957 * E.g. dev1 dev2 dev3 dev4 dev5
961 if (__is_raid10_near(rs
->md
.new_layout
)) {
962 for (i
= 0; i
< rs
->md
.raid_disks
; i
++) {
964 rebuilds_per_group
= 0;
965 if ((!rs
->dev
[i
].rdev
.sb_page
||
966 !test_bit(In_sync
, &rs
->dev
[i
].rdev
.flags
)) &&
967 (++rebuilds_per_group
>= copies
))
974 * When checking "far" and "offset" formats, we need to ensure
975 * that the device that holds its copy is not also dead or
976 * being rebuilt. (Note that "far" and "offset" formats only
977 * support two copies right now. These formats also only ever
978 * use the 'use_far_sets' variant.)
980 * This check is somewhat complicated by the need to account
981 * for arrays that are not a multiple of (far) copies. This
982 * results in the need to treat the last (potentially larger)
985 group_size
= (rs
->md
.raid_disks
/ copies
);
986 last_group_start
= (rs
->md
.raid_disks
/ group_size
) - 1;
987 last_group_start
*= group_size
;
988 for (i
= 0; i
< rs
->md
.raid_disks
; i
++) {
989 if (!(i
% copies
) && !(i
> last_group_start
))
990 rebuilds_per_group
= 0;
991 if ((!rs
->dev
[i
].rdev
.sb_page
||
992 !test_bit(In_sync
, &rs
->dev
[i
].rdev
.flags
)) &&
993 (++rebuilds_per_group
>= copies
))
1009 * Possible arguments are...
1010 * <chunk_size> [optional_args]
1012 * Argument definitions
1013 * <chunk_size> The number of sectors per disk that
1014 * will form the "stripe"
1015 * [[no]sync] Force or prevent recovery of the
1017 * [rebuild <idx>] Rebuild the drive indicated by the index
1018 * [daemon_sleep <ms>] Time between bitmap daemon work to
1020 * [min_recovery_rate <kB/sec/disk>] Throttle RAID initialization
1021 * [max_recovery_rate <kB/sec/disk>] Throttle RAID initialization
1022 * [write_mostly <idx>] Indicate a write mostly drive via index
1023 * [max_write_behind <sectors>] See '-write-behind=' (man mdadm)
1024 * [stripe_cache <sectors>] Stripe cache size for higher RAIDs
1025 * [region_size <sectors>] Defines granularity of bitmap
1027 * RAID10-only options:
1028 * [raid10_copies <# copies>] Number of copies. (Default: 2)
1029 * [raid10_format <near|far|offset>] Layout algorithm. (Default: near)
1031 static int parse_raid_params(struct raid_set
*rs
, struct dm_arg_set
*as
,
1032 unsigned int num_raid_params
)
1034 int value
, raid10_format
= ALGORITHM_RAID10_DEFAULT
;
1035 unsigned int raid10_copies
= 2;
1036 unsigned int i
, write_mostly
= 0;
1037 unsigned int region_size
= 0;
1038 sector_t max_io_len
;
1039 const char *arg
, *key
;
1040 struct raid_dev
*rd
;
1041 struct raid_type
*rt
= rs
->raid_type
;
1043 arg
= dm_shift_arg(as
);
1044 num_raid_params
--; /* Account for chunk_size argument */
1046 if (kstrtoint(arg
, 10, &value
) < 0) {
1047 rs
->ti
->error
= "Bad numerical argument given for chunk_size";
1052 * First, parse the in-order required arguments
1053 * "chunk_size" is the only argument of this type.
1055 if (rt_is_raid1(rt
)) {
1057 DMERR("Ignoring chunk size parameter for RAID 1");
1059 } else if (!is_power_of_2(value
)) {
1060 rs
->ti
->error
= "Chunk size must be a power of 2";
1062 } else if (value
< 8) {
1063 rs
->ti
->error
= "Chunk size value is too small";
1067 rs
->md
.new_chunk_sectors
= rs
->md
.chunk_sectors
= value
;
1070 * We set each individual device as In_sync with a completed
1071 * 'recovery_offset'. If there has been a device failure or
1072 * replacement then one of the following cases applies:
1074 * 1) User specifies 'rebuild'.
1075 * - Device is reset when param is read.
1076 * 2) A new device is supplied.
1077 * - No matching superblock found, resets device.
1078 * 3) Device failure was transient and returns on reload.
1079 * - Failure noticed, resets device for bitmap replay.
1080 * 4) Device hadn't completed recovery after previous failure.
1081 * - Superblock is read and overrides recovery_offset.
1083 * What is found in the superblocks of the devices is always
1084 * authoritative, unless 'rebuild' or '[no]sync' was specified.
1086 for (i
= 0; i
< rs
->raid_disks
; i
++) {
1087 set_bit(In_sync
, &rs
->dev
[i
].rdev
.flags
);
1088 rs
->dev
[i
].rdev
.recovery_offset
= MaxSector
;
1092 * Second, parse the unordered optional arguments
1094 for (i
= 0; i
< num_raid_params
; i
++) {
1095 key
= dm_shift_arg(as
);
1097 rs
->ti
->error
= "Not enough raid parameters given";
1101 if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC
))) {
1102 if (test_and_set_bit(__CTR_FLAG_NOSYNC
, &rs
->ctr_flags
)) {
1103 rs
->ti
->error
= "Only one 'nosync' argument allowed";
1108 if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_SYNC
))) {
1109 if (test_and_set_bit(__CTR_FLAG_SYNC
, &rs
->ctr_flags
)) {
1110 rs
->ti
->error
= "Only one 'sync' argument allowed";
1115 if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_USE_NEAR_SETS
))) {
1116 if (test_and_set_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS
, &rs
->ctr_flags
)) {
1117 rs
->ti
->error
= "Only one 'raid10_use_new_sets' argument allowed";
1123 arg
= dm_shift_arg(as
);
1124 i
++; /* Account for the argument pairs */
1126 rs
->ti
->error
= "Wrong number of raid parameters given";
1131 * Parameters that take a string value are checked here.
1134 if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT
))) {
1135 if (test_and_set_bit(__CTR_FLAG_RAID10_FORMAT
, &rs
->ctr_flags
)) {
1136 rs
->ti
->error
= "Only one 'raid10_format' argument pair allowed";
1139 if (!rt_is_raid10(rt
)) {
1140 rs
->ti
->error
= "'raid10_format' is an invalid parameter for this RAID type";
1143 raid10_format
= raid10_name_to_format(arg
);
1144 if (raid10_format
< 0) {
1145 rs
->ti
->error
= "Invalid 'raid10_format' value given";
1146 return raid10_format
;
1151 if (kstrtoint(arg
, 10, &value
) < 0) {
1152 rs
->ti
->error
= "Bad numerical argument given in raid params";
1156 if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD
))) {
1158 * "rebuild" is being passed in by userspace to provide
1159 * indexes of replaced devices and to set up additional
1160 * devices on raid level takeover.
1162 if (!__within_range(value
, 0, rs
->raid_disks
- 1)) {
1163 rs
->ti
->error
= "Invalid rebuild index given";
1167 if (test_and_set_bit(value
, (void *) rs
->rebuild_disks
)) {
1168 rs
->ti
->error
= "rebuild for this index already given";
1172 rd
= rs
->dev
+ value
;
1173 clear_bit(In_sync
, &rd
->rdev
.flags
);
1174 clear_bit(Faulty
, &rd
->rdev
.flags
);
1175 rd
->rdev
.recovery_offset
= 0;
1176 set_bit(__CTR_FLAG_REBUILD
, &rs
->ctr_flags
);
1177 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY
))) {
1178 if (!rt_is_raid1(rt
)) {
1179 rs
->ti
->error
= "write_mostly option is only valid for RAID1";
1183 if (!__within_range(value
, 0, rs
->md
.raid_disks
- 1)) {
1184 rs
->ti
->error
= "Invalid write_mostly index given";
1189 set_bit(WriteMostly
, &rs
->dev
[value
].rdev
.flags
);
1190 set_bit(__CTR_FLAG_WRITE_MOSTLY
, &rs
->ctr_flags
);
1191 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND
))) {
1192 if (!rt_is_raid1(rt
)) {
1193 rs
->ti
->error
= "max_write_behind option is only valid for RAID1";
1197 if (test_and_set_bit(__CTR_FLAG_MAX_WRITE_BEHIND
, &rs
->ctr_flags
)) {
1198 rs
->ti
->error
= "Only one max_write_behind argument pair allowed";
1203 * In device-mapper, we specify things in sectors, but
1204 * MD records this value in kB
1207 if (value
> COUNTER_MAX
) {
1208 rs
->ti
->error
= "Max write-behind limit out of range";
1212 rs
->md
.bitmap_info
.max_write_behind
= value
;
1213 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP
))) {
1214 if (test_and_set_bit(__CTR_FLAG_DAEMON_SLEEP
, &rs
->ctr_flags
)) {
1215 rs
->ti
->error
= "Only one daemon_sleep argument pair allowed";
1218 if (!value
|| (value
> MAX_SCHEDULE_TIMEOUT
)) {
1219 rs
->ti
->error
= "daemon sleep period out of range";
1222 rs
->md
.bitmap_info
.daemon_sleep
= value
;
1223 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET
))) {
1224 /* Userspace passes new data_offset after having extended the the data image LV */
1225 if (test_and_set_bit(__CTR_FLAG_DATA_OFFSET
, &rs
->ctr_flags
)) {
1226 rs
->ti
->error
= "Only one data_offset argument pair allowed";
1229 /* Ensure sensible data offset */
1231 (value
&& (value
< MIN_FREE_RESHAPE_SPACE
|| value
% to_sector(PAGE_SIZE
)))) {
1232 rs
->ti
->error
= "Bogus data_offset value";
1235 rs
->data_offset
= value
;
1236 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS
))) {
1237 /* Define the +/-# of disks to add to/remove from the given raid set */
1238 if (test_and_set_bit(__CTR_FLAG_DELTA_DISKS
, &rs
->ctr_flags
)) {
1239 rs
->ti
->error
= "Only one delta_disks argument pair allowed";
1242 /* Ensure MAX_RAID_DEVICES and raid type minimal_devs! */
1243 if (!__within_range(abs(value
), 1, MAX_RAID_DEVICES
- rt
->minimal_devs
)) {
1244 rs
->ti
->error
= "Too many delta_disk requested";
1248 rs
->delta_disks
= value
;
1249 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE
))) {
1250 if (test_and_set_bit(__CTR_FLAG_STRIPE_CACHE
, &rs
->ctr_flags
)) {
1251 rs
->ti
->error
= "Only one stripe_cache argument pair allowed";
1255 if (!rt_is_raid456(rt
)) {
1256 rs
->ti
->error
= "Inappropriate argument: stripe_cache";
1260 rs
->stripe_cache_entries
= value
;
1261 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE
))) {
1262 if (test_and_set_bit(__CTR_FLAG_MIN_RECOVERY_RATE
, &rs
->ctr_flags
)) {
1263 rs
->ti
->error
= "Only one min_recovery_rate argument pair allowed";
1266 if (value
> INT_MAX
) {
1267 rs
->ti
->error
= "min_recovery_rate out of range";
1270 rs
->md
.sync_speed_min
= (int)value
;
1271 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE
))) {
1272 if (test_and_set_bit(__CTR_FLAG_MAX_RECOVERY_RATE
, &rs
->ctr_flags
)) {
1273 rs
->ti
->error
= "Only one max_recovery_rate argument pair allowed";
1276 if (value
> INT_MAX
) {
1277 rs
->ti
->error
= "max_recovery_rate out of range";
1280 rs
->md
.sync_speed_max
= (int)value
;
1281 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE
))) {
1282 if (test_and_set_bit(__CTR_FLAG_REGION_SIZE
, &rs
->ctr_flags
)) {
1283 rs
->ti
->error
= "Only one region_size argument pair allowed";
1287 region_size
= value
;
1288 rs
->requested_bitmap_chunk_sectors
= value
;
1289 } else if (!strcasecmp(key
, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES
))) {
1290 if (test_and_set_bit(__CTR_FLAG_RAID10_COPIES
, &rs
->ctr_flags
)) {
1291 rs
->ti
->error
= "Only one raid10_copies argument pair allowed";
1295 if (!__within_range(value
, 2, rs
->md
.raid_disks
)) {
1296 rs
->ti
->error
= "Bad value for 'raid10_copies'";
1300 raid10_copies
= value
;
1302 DMERR("Unable to parse RAID parameter: %s", key
);
1303 rs
->ti
->error
= "Unable to parse RAID parameter";
1308 if (test_bit(__CTR_FLAG_SYNC
, &rs
->ctr_flags
) &&
1309 test_bit(__CTR_FLAG_NOSYNC
, &rs
->ctr_flags
)) {
1310 rs
->ti
->error
= "sync and nosync are mutually exclusive";
1314 if (test_bit(__CTR_FLAG_REBUILD
, &rs
->ctr_flags
) &&
1315 (test_bit(__CTR_FLAG_SYNC
, &rs
->ctr_flags
) ||
1316 test_bit(__CTR_FLAG_NOSYNC
, &rs
->ctr_flags
))) {
1317 rs
->ti
->error
= "sync/nosync and rebuild are mutually exclusive";
1321 if (write_mostly
>= rs
->md
.raid_disks
) {
1322 rs
->ti
->error
= "Can't set all raid1 devices to write_mostly";
1326 if (validate_region_size(rs
, region_size
))
1329 if (rs
->md
.chunk_sectors
)
1330 max_io_len
= rs
->md
.chunk_sectors
;
1332 max_io_len
= region_size
;
1334 if (dm_set_target_max_io_len(rs
->ti
, max_io_len
))
1337 if (rt_is_raid10(rt
)) {
1338 if (raid10_copies
> rs
->md
.raid_disks
) {
1339 rs
->ti
->error
= "Not enough devices to satisfy specification";
1343 rs
->md
.new_layout
= raid10_format_to_md_layout(rs
, raid10_format
, raid10_copies
);
1344 if (rs
->md
.new_layout
< 0) {
1345 rs
->ti
->error
= "Error getting raid10 format";
1346 return rs
->md
.new_layout
;
1349 rt
= get_raid_type_by_ll(10, rs
->md
.new_layout
);
1351 rs
->ti
->error
= "Failed to recognize new raid10 layout";
1355 if ((rt
->algorithm
== ALGORITHM_RAID10_DEFAULT
||
1356 rt
->algorithm
== ALGORITHM_RAID10_NEAR
) &&
1357 test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS
, &rs
->ctr_flags
)) {
1358 rs
->ti
->error
= "RAID10 format 'near' and 'raid10_use_near_sets' are incompatible";
1363 rs
->raid10_copies
= raid10_copies
;
1365 /* Assume there are no metadata devices until the drives are parsed */
1366 rs
->md
.persistent
= 0;
1367 rs
->md
.external
= 1;
1369 /* Check, if any invalid ctr arguments have been passed in for the raid level */
1370 return rs_check_for_valid_flags(rs
);
1373 /* Set raid4/5/6 cache size */
1374 static int rs_set_raid456_stripe_cache(struct raid_set
*rs
)
1377 struct r5conf
*conf
;
1378 struct mddev
*mddev
= &rs
->md
;
1379 uint32_t min_stripes
= max(mddev
->chunk_sectors
, mddev
->new_chunk_sectors
) / 2;
1380 uint32_t nr_stripes
= rs
->stripe_cache_entries
;
1382 if (!rt_is_raid456(rs
->raid_type
)) {
1383 rs
->ti
->error
= "Inappropriate raid level; cannot change stripe_cache size";
1387 if (nr_stripes
< min_stripes
) {
1388 DMINFO("Adjusting requested %u stripe cache entries to %u to suit stripe size",
1389 nr_stripes
, min_stripes
);
1390 nr_stripes
= min_stripes
;
1393 conf
= mddev
->private;
1395 rs
->ti
->error
= "Cannot change stripe_cache size on inactive RAID set";
1399 /* Try setting number of stripes in raid456 stripe cache */
1400 if (conf
->min_nr_stripes
!= nr_stripes
) {
1401 r
= raid5_set_cache_size(mddev
, nr_stripes
);
1403 rs
->ti
->error
= "Failed to set raid4/5/6 stripe cache size";
1407 DMINFO("%u stripe cache entries", nr_stripes
);
1413 /* Return # of data stripes as kept in mddev as of @rs (i.e. as of superblock) */
1414 static unsigned int mddev_data_stripes(struct raid_set
*rs
)
1416 return rs
->md
.raid_disks
- rs
->raid_type
->parity_devs
;
1419 /* Return # of data stripes of @rs (i.e. as of ctr) */
1420 static unsigned int rs_data_stripes(struct raid_set
*rs
)
1422 return rs
->raid_disks
- rs
->raid_type
->parity_devs
;
1425 /* Calculate the sectors per device and per array used for @rs */
1426 static int rs_set_dev_and_array_sectors(struct raid_set
*rs
, bool use_mddev
)
1429 unsigned int data_stripes
;
1430 struct mddev
*mddev
= &rs
->md
;
1431 struct md_rdev
*rdev
;
1432 sector_t array_sectors
= rs
->ti
->len
, dev_sectors
= rs
->ti
->len
;
1435 delta_disks
= mddev
->delta_disks
;
1436 data_stripes
= mddev_data_stripes(rs
);
1438 delta_disks
= rs
->delta_disks
;
1439 data_stripes
= rs_data_stripes(rs
);
1442 /* Special raid1 case w/o delta_disks support (yet) */
1443 if (rt_is_raid1(rs
->raid_type
))
1445 else if (rt_is_raid10(rs
->raid_type
)) {
1446 if (rs
->raid10_copies
< 2 ||
1448 rs
->ti
->error
= "Bogus raid10 data copies or delta disks";
1452 dev_sectors
*= rs
->raid10_copies
;
1453 if (sector_div(dev_sectors
, data_stripes
))
1456 array_sectors
= (data_stripes
+ delta_disks
) * dev_sectors
;
1457 if (sector_div(array_sectors
, rs
->raid10_copies
))
1460 } else if (sector_div(dev_sectors
, data_stripes
))
1464 /* Striped layouts */
1465 array_sectors
= (data_stripes
+ delta_disks
) * dev_sectors
;
1467 rdev_for_each(rdev
, mddev
)
1468 rdev
->sectors
= dev_sectors
;
1470 mddev
->array_sectors
= array_sectors
;
1471 mddev
->dev_sectors
= dev_sectors
;
1475 rs
->ti
->error
= "Target length not divisible by number of data devices";
1479 /* Setup recovery on @rs */
1480 static void __rs_setup_recovery(struct raid_set
*rs
, sector_t dev_sectors
)
1482 /* raid0 does not recover */
1483 if (rs_is_raid0(rs
))
1484 rs
->md
.recovery_cp
= MaxSector
;
1486 * A raid6 set has to be recovered either
1487 * completely or for the grown part to
1488 * ensure proper parity and Q-Syndrome
1490 else if (rs_is_raid6(rs
))
1491 rs
->md
.recovery_cp
= dev_sectors
;
1493 * Other raid set types may skip recovery
1494 * depending on the 'nosync' flag.
1497 rs
->md
.recovery_cp
= test_bit(__CTR_FLAG_NOSYNC
, &rs
->ctr_flags
)
1498 ? MaxSector
: dev_sectors
;
1501 /* Setup recovery on @rs based on raid type, device size and 'nosync' flag */
1502 static void rs_setup_recovery(struct raid_set
*rs
, sector_t dev_sectors
)
1505 /* New raid set or 'sync' flag provided */
1506 __rs_setup_recovery(rs
, 0);
1507 else if (dev_sectors
== MaxSector
)
1508 /* Prevent recovery */
1509 __rs_setup_recovery(rs
, MaxSector
);
1510 else if (rs
->dev
[0].rdev
.sectors
< dev_sectors
)
1511 /* Grown raid set */
1512 __rs_setup_recovery(rs
, rs
->dev
[0].rdev
.sectors
);
1514 __rs_setup_recovery(rs
, MaxSector
);
1517 static void do_table_event(struct work_struct
*ws
)
1519 struct raid_set
*rs
= container_of(ws
, struct raid_set
, md
.event_work
);
1521 smp_rmb(); /* Make sure we access most actual mddev properties */
1522 if (!rs_is_reshaping(rs
))
1523 rs_set_capacity(rs
);
1524 dm_table_event(rs
->ti
->table
);
1527 static int raid_is_congested(struct dm_target_callbacks
*cb
, int bits
)
1529 struct raid_set
*rs
= container_of(cb
, struct raid_set
, callbacks
);
1531 return mddev_congested(&rs
->md
, bits
);
1535 * Make sure a valid takover (level switch) is being requested on @rs
1537 * Conversions of raid sets from one MD personality to another
1538 * have to conform to restrictions which are enforced here.
1540 static int rs_check_takeover(struct raid_set
*rs
)
1542 struct mddev
*mddev
= &rs
->md
;
1543 unsigned int near_copies
;
1545 if (rs
->md
.degraded
) {
1546 rs
->ti
->error
= "Can't takeover degraded raid set";
1550 if (rs_is_reshaping(rs
)) {
1551 rs
->ti
->error
= "Can't takeover reshaping raid set";
1555 switch (mddev
->level
) {
1557 /* raid0 -> raid1/5 with one disk */
1558 if ((mddev
->new_level
== 1 || mddev
->new_level
== 5) &&
1559 mddev
->raid_disks
== 1)
1562 /* raid0 -> raid10 */
1563 if (mddev
->new_level
== 10 &&
1564 !(rs
->raid_disks
% mddev
->raid_disks
))
1567 /* raid0 with multiple disks -> raid4/5/6 */
1568 if (__within_range(mddev
->new_level
, 4, 6) &&
1569 mddev
->new_layout
== ALGORITHM_PARITY_N
&&
1570 mddev
->raid_disks
> 1)
1576 /* Can't takeover raid10_offset! */
1577 if (__is_raid10_offset(mddev
->layout
))
1580 near_copies
= __raid10_near_copies(mddev
->layout
);
1582 /* raid10* -> raid0 */
1583 if (mddev
->new_level
== 0) {
1584 /* Can takeover raid10_near with raid disks divisable by data copies! */
1585 if (near_copies
> 1 &&
1586 !(mddev
->raid_disks
% near_copies
)) {
1587 mddev
->raid_disks
/= near_copies
;
1588 mddev
->delta_disks
= mddev
->raid_disks
;
1592 /* Can takeover raid10_far */
1593 if (near_copies
== 1 &&
1594 __raid10_far_copies(mddev
->layout
) > 1)
1600 /* raid10_{near,far} -> raid1 */
1601 if (mddev
->new_level
== 1 &&
1602 max(near_copies
, __raid10_far_copies(mddev
->layout
)) == mddev
->raid_disks
)
1605 /* raid10_{near,far} with 2 disks -> raid4/5 */
1606 if (__within_range(mddev
->new_level
, 4, 5) &&
1607 mddev
->raid_disks
== 2)
1612 /* raid1 with 2 disks -> raid4/5 */
1613 if (__within_range(mddev
->new_level
, 4, 5) &&
1614 mddev
->raid_disks
== 2) {
1615 mddev
->degraded
= 1;
1619 /* raid1 -> raid0 */
1620 if (mddev
->new_level
== 0 &&
1621 mddev
->raid_disks
== 1)
1624 /* raid1 -> raid10 */
1625 if (mddev
->new_level
== 10)
1630 /* raid4 -> raid0 */
1631 if (mddev
->new_level
== 0)
1634 /* raid4 -> raid1/5 with 2 disks */
1635 if ((mddev
->new_level
== 1 || mddev
->new_level
== 5) &&
1636 mddev
->raid_disks
== 2)
1639 /* raid4 -> raid5/6 with parity N */
1640 if (__within_range(mddev
->new_level
, 5, 6) &&
1641 mddev
->layout
== ALGORITHM_PARITY_N
)
1646 /* raid5 with parity N -> raid0 */
1647 if (mddev
->new_level
== 0 &&
1648 mddev
->layout
== ALGORITHM_PARITY_N
)
1651 /* raid5 with parity N -> raid4 */
1652 if (mddev
->new_level
== 4 &&
1653 mddev
->layout
== ALGORITHM_PARITY_N
)
1656 /* raid5 with 2 disks -> raid1/4/10 */
1657 if ((mddev
->new_level
== 1 || mddev
->new_level
== 4 || mddev
->new_level
== 10) &&
1658 mddev
->raid_disks
== 2)
1661 /* raid5_* -> raid6_*_6 with Q-Syndrome N (e.g. raid5_ra -> raid6_ra_6 */
1662 if (mddev
->new_level
== 6 &&
1663 ((mddev
->layout
== ALGORITHM_PARITY_N
&& mddev
->new_layout
== ALGORITHM_PARITY_N
) ||
1664 __within_range(mddev
->new_layout
, ALGORITHM_LEFT_ASYMMETRIC_6
, ALGORITHM_RIGHT_SYMMETRIC_6
)))
1669 /* raid6 with parity N -> raid0 */
1670 if (mddev
->new_level
== 0 &&
1671 mddev
->layout
== ALGORITHM_PARITY_N
)
1674 /* raid6 with parity N -> raid4 */
1675 if (mddev
->new_level
== 4 &&
1676 mddev
->layout
== ALGORITHM_PARITY_N
)
1679 /* raid6_*_n with Q-Syndrome N -> raid5_* */
1680 if (mddev
->new_level
== 5 &&
1681 ((mddev
->layout
== ALGORITHM_PARITY_N
&& mddev
->new_layout
== ALGORITHM_PARITY_N
) ||
1682 __within_range(mddev
->new_layout
, ALGORITHM_LEFT_ASYMMETRIC
, ALGORITHM_RIGHT_SYMMETRIC
)))
1689 rs
->ti
->error
= "takeover not possible";
1693 /* True if @rs requested to be taken over */
1694 static bool rs_takeover_requested(struct raid_set
*rs
)
1696 return rs
->md
.new_level
!= rs
->md
.level
;
1699 /* True if @rs is requested to reshape by ctr */
1700 static bool rs_reshape_requested(struct raid_set
*rs
)
1703 struct mddev
*mddev
= &rs
->md
;
1705 if (rs_takeover_requested(rs
))
1711 change
= mddev
->new_layout
!= mddev
->layout
||
1712 mddev
->new_chunk_sectors
!= mddev
->chunk_sectors
||
1715 /* Historical case to support raid1 reshape without delta disks */
1716 if (mddev
->level
== 1) {
1717 if (rs
->delta_disks
)
1718 return !!rs
->delta_disks
;
1721 mddev
->raid_disks
!= rs
->raid_disks
;
1724 if (mddev
->level
== 10)
1726 !__is_raid10_far(mddev
->new_layout
) &&
1727 rs
->delta_disks
>= 0;
1733 #define FEATURE_FLAG_SUPPORTS_V190 0x1 /* Supports extended superblock */
1735 /* State flags for sb->flags */
1736 #define SB_FLAG_RESHAPE_ACTIVE 0x1
1737 #define SB_FLAG_RESHAPE_BACKWARDS 0x2
1740 * This structure is never routinely used by userspace, unlike md superblocks.
1741 * Devices with this superblock should only ever be accessed via device-mapper.
1743 #define DM_RAID_MAGIC 0x64526D44
1744 struct dm_raid_superblock
{
1745 __le32 magic
; /* "DmRd" */
1746 __le32 compat_features
; /* Used to indicate compatible features (like 1.9.0 ondisk metadata extension) */
1748 __le32 num_devices
; /* Number of devices in this raid set. (Max 64) */
1749 __le32 array_position
; /* The position of this drive in the raid set */
1751 __le64 events
; /* Incremented by md when superblock updated */
1752 __le64 failed_devices
; /* Pre 1.9.0 part of bit field of devices to */
1753 /* indicate failures (see extension below) */
1756 * This offset tracks the progress of the repair or replacement of
1757 * an individual drive.
1759 __le64 disk_recovery_offset
;
1762 * This offset tracks the progress of the initial raid set
1763 * synchronisation/parity calculation.
1765 __le64 array_resync_offset
;
1768 * raid characteristics
1772 __le32 stripe_sectors
;
1774 /********************************************************************
1775 * BELOW FOLLOW V1.9.0 EXTENSIONS TO THE PRISTINE SUPERBLOCK FORMAT!!!
1777 * FEATURE_FLAG_SUPPORTS_V190 in the features member indicates that those exist
1780 __le32 flags
; /* Flags defining array states for reshaping */
1783 * This offset tracks the progress of a raid
1784 * set reshape in order to be able to restart it
1786 __le64 reshape_position
;
1789 * These define the properties of the array in case of an interrupted reshape
1793 __le32 new_stripe_sectors
;
1796 __le64 array_sectors
; /* Array size in sectors */
1799 * Sector offsets to data on devices (reshaping).
1800 * Needed to support out of place reshaping, thus
1801 * not writing over any stripes whilst converting
1802 * them from old to new layout
1805 __le64 new_data_offset
;
1807 __le64 sectors
; /* Used device size in sectors */
1810 * Additonal Bit field of devices indicating failures to support
1811 * up to 256 devices with the 1.9.0 on-disk metadata format
1813 __le64 extended_failed_devices
[DISKS_ARRAY_ELEMS
- 1];
1815 __le32 incompat_features
; /* Used to indicate any incompatible features */
1817 /* Always set rest up to logical block size to 0 when writing (see get_metadata_device() below). */
1821 * Check for reshape constraints on raid set @rs:
1823 * - reshape function non-existent
1825 * - ongoing recovery
1828 * Returns 0 if none or -EPERM if given constraint
1829 * and error message reference in @errmsg
1831 static int rs_check_reshape(struct raid_set
*rs
)
1833 struct mddev
*mddev
= &rs
->md
;
1835 if (!mddev
->pers
|| !mddev
->pers
->check_reshape
)
1836 rs
->ti
->error
= "Reshape not supported";
1837 else if (mddev
->degraded
)
1838 rs
->ti
->error
= "Can't reshape degraded raid set";
1839 else if (rs_is_recovering(rs
))
1840 rs
->ti
->error
= "Convert request on recovering raid set prohibited";
1841 else if (rs_is_reshaping(rs
))
1842 rs
->ti
->error
= "raid set already reshaping!";
1843 else if (!(rs_is_raid1(rs
) || rs_is_raid10(rs
) || rs_is_raid456(rs
)))
1844 rs
->ti
->error
= "Reshaping only supported for raid1/4/5/6/10";
1851 static int read_disk_sb(struct md_rdev
*rdev
, int size
)
1853 BUG_ON(!rdev
->sb_page
);
1855 if (rdev
->sb_loaded
)
1858 if (!sync_page_io(rdev
, 0, size
, rdev
->sb_page
, REQ_OP_READ
, 0, true)) {
1859 DMERR("Failed to read superblock of device at position %d",
1861 md_error(rdev
->mddev
, rdev
);
1865 rdev
->sb_loaded
= 1;
1870 static void sb_retrieve_failed_devices(struct dm_raid_superblock
*sb
, uint64_t *failed_devices
)
1872 failed_devices
[0] = le64_to_cpu(sb
->failed_devices
);
1873 memset(failed_devices
+ 1, 0, sizeof(sb
->extended_failed_devices
));
1875 if (le32_to_cpu(sb
->compat_features
) & FEATURE_FLAG_SUPPORTS_V190
) {
1876 int i
= ARRAY_SIZE(sb
->extended_failed_devices
);
1879 failed_devices
[i
+1] = le64_to_cpu(sb
->extended_failed_devices
[i
]);
1883 static void sb_update_failed_devices(struct dm_raid_superblock
*sb
, uint64_t *failed_devices
)
1885 int i
= ARRAY_SIZE(sb
->extended_failed_devices
);
1887 sb
->failed_devices
= cpu_to_le64(failed_devices
[0]);
1889 sb
->extended_failed_devices
[i
] = cpu_to_le64(failed_devices
[i
+1]);
1893 * Synchronize the superblock members with the raid set properties
1895 * All superblock data is little endian.
1897 static void super_sync(struct mddev
*mddev
, struct md_rdev
*rdev
)
1899 bool update_failed_devices
= false;
1901 uint64_t failed_devices
[DISKS_ARRAY_ELEMS
];
1902 struct dm_raid_superblock
*sb
;
1903 struct raid_set
*rs
= container_of(mddev
, struct raid_set
, md
);
1905 /* No metadata device, no superblock */
1906 if (!rdev
->meta_bdev
)
1909 BUG_ON(!rdev
->sb_page
);
1911 sb
= page_address(rdev
->sb_page
);
1913 sb_retrieve_failed_devices(sb
, failed_devices
);
1915 for (i
= 0; i
< rs
->raid_disks
; i
++)
1916 if (!rs
->dev
[i
].data_dev
|| test_bit(Faulty
, &rs
->dev
[i
].rdev
.flags
)) {
1917 update_failed_devices
= true;
1918 set_bit(i
, (void *) failed_devices
);
1921 if (update_failed_devices
)
1922 sb_update_failed_devices(sb
, failed_devices
);
1924 sb
->magic
= cpu_to_le32(DM_RAID_MAGIC
);
1925 sb
->compat_features
= cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190
);
1927 sb
->num_devices
= cpu_to_le32(mddev
->raid_disks
);
1928 sb
->array_position
= cpu_to_le32(rdev
->raid_disk
);
1930 sb
->events
= cpu_to_le64(mddev
->events
);
1932 sb
->disk_recovery_offset
= cpu_to_le64(rdev
->recovery_offset
);
1933 sb
->array_resync_offset
= cpu_to_le64(mddev
->recovery_cp
);
1935 sb
->level
= cpu_to_le32(mddev
->level
);
1936 sb
->layout
= cpu_to_le32(mddev
->layout
);
1937 sb
->stripe_sectors
= cpu_to_le32(mddev
->chunk_sectors
);
1939 sb
->new_level
= cpu_to_le32(mddev
->new_level
);
1940 sb
->new_layout
= cpu_to_le32(mddev
->new_layout
);
1941 sb
->new_stripe_sectors
= cpu_to_le32(mddev
->new_chunk_sectors
);
1943 sb
->delta_disks
= cpu_to_le32(mddev
->delta_disks
);
1945 smp_rmb(); /* Make sure we access most recent reshape position */
1946 sb
->reshape_position
= cpu_to_le64(mddev
->reshape_position
);
1947 if (le64_to_cpu(sb
->reshape_position
) != MaxSector
) {
1948 /* Flag ongoing reshape */
1949 sb
->flags
|= cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE
);
1951 if (mddev
->delta_disks
< 0 || mddev
->reshape_backwards
)
1952 sb
->flags
|= cpu_to_le32(SB_FLAG_RESHAPE_BACKWARDS
);
1954 /* Clear reshape flags */
1955 sb
->flags
&= ~(cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE
|SB_FLAG_RESHAPE_BACKWARDS
));
1958 sb
->array_sectors
= cpu_to_le64(mddev
->array_sectors
);
1959 sb
->data_offset
= cpu_to_le64(rdev
->data_offset
);
1960 sb
->new_data_offset
= cpu_to_le64(rdev
->new_data_offset
);
1961 sb
->sectors
= cpu_to_le64(rdev
->sectors
);
1962 sb
->incompat_features
= cpu_to_le32(0);
1964 /* Zero out the rest of the payload after the size of the superblock */
1965 memset(sb
+ 1, 0, rdev
->sb_size
- sizeof(*sb
));
1971 * This function creates a superblock if one is not found on the device
1972 * and will decide which superblock to use if there's a choice.
1974 * Return: 1 if use rdev, 0 if use refdev, -Exxx otherwise
1976 static int super_load(struct md_rdev
*rdev
, struct md_rdev
*refdev
)
1979 struct dm_raid_superblock
*sb
;
1980 struct dm_raid_superblock
*refsb
;
1981 uint64_t events_sb
, events_refsb
;
1984 rdev
->sb_size
= bdev_logical_block_size(rdev
->meta_bdev
);
1985 if (rdev
->sb_size
< sizeof(*sb
) || rdev
->sb_size
> PAGE_SIZE
) {
1986 DMERR("superblock size of a logical block is no longer valid");
1990 r
= read_disk_sb(rdev
, rdev
->sb_size
);
1994 sb
= page_address(rdev
->sb_page
);
1997 * Two cases that we want to write new superblocks and rebuild:
1998 * 1) New device (no matching magic number)
1999 * 2) Device specified for rebuild (!In_sync w/ offset == 0)
2001 if ((sb
->magic
!= cpu_to_le32(DM_RAID_MAGIC
)) ||
2002 (!test_bit(In_sync
, &rdev
->flags
) && !rdev
->recovery_offset
)) {
2003 super_sync(rdev
->mddev
, rdev
);
2005 set_bit(FirstUse
, &rdev
->flags
);
2006 sb
->compat_features
= cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190
);
2008 /* Force writing of superblocks to disk */
2009 set_bit(MD_CHANGE_DEVS
, &rdev
->mddev
->flags
);
2011 /* Any superblock is better than none, choose that if given */
2012 return refdev
? 0 : 1;
2018 events_sb
= le64_to_cpu(sb
->events
);
2020 refsb
= page_address(refdev
->sb_page
);
2021 events_refsb
= le64_to_cpu(refsb
->events
);
2023 return (events_sb
> events_refsb
) ? 1 : 0;
2026 static int super_init_validation(struct raid_set
*rs
, struct md_rdev
*rdev
)
2030 struct mddev
*mddev
= &rs
->md
;
2032 uint64_t failed_devices
[DISKS_ARRAY_ELEMS
];
2033 struct dm_raid_superblock
*sb
;
2034 uint32_t new_devs
= 0, rebuild_and_new
= 0, rebuilds
= 0;
2036 struct dm_raid_superblock
*sb2
;
2038 sb
= page_address(rdev
->sb_page
);
2039 events_sb
= le64_to_cpu(sb
->events
);
2042 * Initialise to 1 if this is a new superblock.
2044 mddev
->events
= events_sb
? : 1;
2046 mddev
->reshape_position
= MaxSector
;
2049 * Reshaping is supported, e.g. reshape_position is valid
2050 * in superblock and superblock content is authoritative.
2052 if (le32_to_cpu(sb
->compat_features
) & FEATURE_FLAG_SUPPORTS_V190
) {
2053 /* Superblock is authoritative wrt given raid set layout! */
2054 mddev
->raid_disks
= le32_to_cpu(sb
->num_devices
);
2055 mddev
->level
= le32_to_cpu(sb
->level
);
2056 mddev
->layout
= le32_to_cpu(sb
->layout
);
2057 mddev
->chunk_sectors
= le32_to_cpu(sb
->stripe_sectors
);
2058 mddev
->new_level
= le32_to_cpu(sb
->new_level
);
2059 mddev
->new_layout
= le32_to_cpu(sb
->new_layout
);
2060 mddev
->new_chunk_sectors
= le32_to_cpu(sb
->new_stripe_sectors
);
2061 mddev
->delta_disks
= le32_to_cpu(sb
->delta_disks
);
2062 mddev
->array_sectors
= le64_to_cpu(sb
->array_sectors
);
2064 /* raid was reshaping and got interrupted */
2065 if (le32_to_cpu(sb
->flags
) & SB_FLAG_RESHAPE_ACTIVE
) {
2066 if (test_bit(__CTR_FLAG_DELTA_DISKS
, &rs
->ctr_flags
)) {
2067 DMERR("Reshape requested but raid set is still reshaping");
2071 if (mddev
->delta_disks
< 0 ||
2072 (!mddev
->delta_disks
&& (le32_to_cpu(sb
->flags
) & SB_FLAG_RESHAPE_BACKWARDS
)))
2073 mddev
->reshape_backwards
= 1;
2075 mddev
->reshape_backwards
= 0;
2077 mddev
->reshape_position
= le64_to_cpu(sb
->reshape_position
);
2078 rs
->raid_type
= get_raid_type_by_ll(mddev
->level
, mddev
->layout
);
2083 * No takeover/reshaping, because we don't have the extended v1.9.0 metadata
2085 if (le32_to_cpu(sb
->level
) != mddev
->level
) {
2086 DMERR("Reshaping/takeover raid sets not yet supported. (raid level/stripes/size change)");
2089 if (le32_to_cpu(sb
->layout
) != mddev
->layout
) {
2090 DMERR("Reshaping raid sets not yet supported. (raid layout change)");
2091 DMERR(" 0x%X vs 0x%X", le32_to_cpu(sb
->layout
), mddev
->layout
);
2092 DMERR(" Old layout: %s w/ %d copies",
2093 raid10_md_layout_to_format(le32_to_cpu(sb
->layout
)),
2094 raid10_md_layout_to_copies(le32_to_cpu(sb
->layout
)));
2095 DMERR(" New layout: %s w/ %d copies",
2096 raid10_md_layout_to_format(mddev
->layout
),
2097 raid10_md_layout_to_copies(mddev
->layout
));
2100 if (le32_to_cpu(sb
->stripe_sectors
) != mddev
->chunk_sectors
) {
2101 DMERR("Reshaping raid sets not yet supported. (stripe sectors change)");
2105 /* We can only change the number of devices in raid1 with old (i.e. pre 1.0.7) metadata */
2106 if (!rt_is_raid1(rs
->raid_type
) &&
2107 (le32_to_cpu(sb
->num_devices
) != mddev
->raid_disks
)) {
2108 DMERR("Reshaping raid sets not yet supported. (device count change from %u to %u)",
2109 sb
->num_devices
, mddev
->raid_disks
);
2113 /* Table line is checked vs. authoritative superblock */
2117 if (!test_bit(__CTR_FLAG_NOSYNC
, &rs
->ctr_flags
))
2118 mddev
->recovery_cp
= le64_to_cpu(sb
->array_resync_offset
);
2121 * During load, we set FirstUse if a new superblock was written.
2122 * There are two reasons we might not have a superblock:
2123 * 1) The raid set is brand new - in which case, all of the
2124 * devices must have their In_sync bit set. Also,
2125 * recovery_cp must be 0, unless forced.
2126 * 2) This is a new device being added to an old raid set
2127 * and the new device needs to be rebuilt - in which
2128 * case the In_sync bit will /not/ be set and
2129 * recovery_cp must be MaxSector.
2130 * 3) This is/are a new device(s) being added to an old
2131 * raid set during takeover to a higher raid level
2132 * to provide capacity for redundancy or during reshape
2133 * to add capacity to grow the raid set.
2136 rdev_for_each(r
, mddev
) {
2137 if (test_bit(FirstUse
, &r
->flags
))
2140 if (!test_bit(In_sync
, &r
->flags
)) {
2141 DMINFO("Device %d specified for rebuild; clearing superblock",
2145 if (test_bit(FirstUse
, &r
->flags
))
2152 if (new_devs
== rs
->raid_disks
|| !rebuilds
) {
2153 /* Replace a broken device */
2154 if (new_devs
== 1 && !rs
->delta_disks
)
2156 if (new_devs
== rs
->raid_disks
) {
2157 DMINFO("Superblocks created for new raid set");
2158 set_bit(MD_ARRAY_FIRST_USE
, &mddev
->flags
);
2159 } else if (new_devs
!= rebuilds
&&
2160 new_devs
!= rs
->delta_disks
) {
2161 DMERR("New device injected into existing raid set without "
2162 "'delta_disks' or 'rebuild' parameter specified");
2165 } else if (new_devs
&& new_devs
!= rebuilds
) {
2166 DMERR("%u 'rebuild' devices cannot be injected into"
2167 " a raid set with %u other first-time devices",
2168 rebuilds
, new_devs
);
2170 } else if (rebuilds
) {
2171 if (rebuild_and_new
&& rebuilds
!= rebuild_and_new
) {
2172 DMERR("new device%s provided without 'rebuild'",
2173 new_devs
> 1 ? "s" : "");
2175 } else if (rs_is_recovering(rs
)) {
2176 DMERR("'rebuild' specified while raid set is not in-sync (recovery_cp=%llu)",
2177 (unsigned long long) mddev
->recovery_cp
);
2179 } else if (rs_is_reshaping(rs
)) {
2180 DMERR("'rebuild' specified while raid set is being reshaped (reshape_position=%llu)",
2181 (unsigned long long) mddev
->reshape_position
);
2187 * Now we set the Faulty bit for those devices that are
2188 * recorded in the superblock as failed.
2190 sb_retrieve_failed_devices(sb
, failed_devices
);
2191 rdev_for_each(r
, mddev
) {
2194 sb2
= page_address(r
->sb_page
);
2195 sb2
->failed_devices
= 0;
2196 memset(sb2
->extended_failed_devices
, 0, sizeof(sb2
->extended_failed_devices
));
2199 * Check for any device re-ordering.
2201 if (!test_bit(FirstUse
, &r
->flags
) && (r
->raid_disk
>= 0)) {
2202 role
= le32_to_cpu(sb2
->array_position
);
2206 if (role
!= r
->raid_disk
) {
2207 if (__is_raid10_near(mddev
->layout
)) {
2208 if (mddev
->raid_disks
% __raid10_near_copies(mddev
->layout
) ||
2209 rs
->raid_disks
% rs
->raid10_copies
) {
2211 "Cannot change raid10 near set to odd # of devices!";
2215 sb2
->array_position
= cpu_to_le32(r
->raid_disk
);
2217 } else if (!(rs_is_raid10(rs
) && rt_is_raid0(rs
->raid_type
)) &&
2218 !(rs_is_raid0(rs
) && rt_is_raid10(rs
->raid_type
)) &&
2219 !rt_is_raid1(rs
->raid_type
)) {
2220 rs
->ti
->error
= "Cannot change device positions in raid set";
2224 DMINFO("raid device #%d now at position #%d", role
, r
->raid_disk
);
2228 * Partial recovery is performed on
2229 * returning failed devices.
2231 if (test_bit(role
, (void *) failed_devices
))
2232 set_bit(Faulty
, &r
->flags
);
2239 static int super_validate(struct raid_set
*rs
, struct md_rdev
*rdev
)
2241 struct mddev
*mddev
= &rs
->md
;
2242 struct dm_raid_superblock
*sb
;
2244 if (rs_is_raid0(rs
) || !rdev
->sb_page
)
2247 sb
= page_address(rdev
->sb_page
);
2250 * If mddev->events is not set, we know we have not yet initialized
2253 if (!mddev
->events
&& super_init_validation(rs
, rdev
))
2256 if (le32_to_cpu(sb
->compat_features
) != FEATURE_FLAG_SUPPORTS_V190
) {
2257 rs
->ti
->error
= "Unable to assemble array: Unknown flag(s) in compatible feature flags";
2261 if (sb
->incompat_features
) {
2262 rs
->ti
->error
= "Unable to assemble array: No incompatible feature flags supported yet";
2266 /* Enable bitmap creation for RAID levels != 0 */
2267 mddev
->bitmap_info
.offset
= rt_is_raid0(rs
->raid_type
) ? 0 : to_sector(4096);
2268 rdev
->mddev
->bitmap_info
.default_offset
= mddev
->bitmap_info
.offset
;
2270 if (!test_and_clear_bit(FirstUse
, &rdev
->flags
)) {
2271 /* Retrieve device size stored in superblock to be prepared for shrink */
2272 rdev
->sectors
= le64_to_cpu(sb
->sectors
);
2273 rdev
->recovery_offset
= le64_to_cpu(sb
->disk_recovery_offset
);
2274 if (rdev
->recovery_offset
== MaxSector
)
2275 set_bit(In_sync
, &rdev
->flags
);
2277 * If no reshape in progress -> we're recovering single
2278 * disk(s) and have to set the device(s) to out-of-sync
2280 else if (!rs_is_reshaping(rs
))
2281 clear_bit(In_sync
, &rdev
->flags
); /* Mandatory for recovery */
2285 * If a device comes back, set it as not In_sync and no longer faulty.
2287 if (test_and_clear_bit(Faulty
, &rdev
->flags
)) {
2288 rdev
->recovery_offset
= 0;
2289 clear_bit(In_sync
, &rdev
->flags
);
2290 rdev
->saved_raid_disk
= rdev
->raid_disk
;
2293 /* Reshape support -> restore repective data offsets */
2294 rdev
->data_offset
= le64_to_cpu(sb
->data_offset
);
2295 rdev
->new_data_offset
= le64_to_cpu(sb
->new_data_offset
);
2301 * Analyse superblocks and select the freshest.
2303 static int analyse_superblocks(struct dm_target
*ti
, struct raid_set
*rs
)
2306 struct raid_dev
*dev
;
2307 struct md_rdev
*rdev
, *tmp
, *freshest
;
2308 struct mddev
*mddev
= &rs
->md
;
2311 rdev_for_each_safe(rdev
, tmp
, mddev
) {
2313 * Skipping super_load due to CTR_FLAG_SYNC will cause
2314 * the array to undergo initialization again as
2315 * though it were new. This is the intended effect
2316 * of the "sync" directive.
2318 * When reshaping capability is added, we must ensure
2319 * that the "sync" directive is disallowed during the
2322 if (test_bit(__CTR_FLAG_SYNC
, &rs
->ctr_flags
))
2325 if (!rdev
->meta_bdev
)
2328 r
= super_load(rdev
, freshest
);
2337 dev
= container_of(rdev
, struct raid_dev
, rdev
);
2339 dm_put_device(ti
, dev
->meta_dev
);
2341 dev
->meta_dev
= NULL
;
2342 rdev
->meta_bdev
= NULL
;
2345 put_page(rdev
->sb_page
);
2347 rdev
->sb_page
= NULL
;
2349 rdev
->sb_loaded
= 0;
2352 * We might be able to salvage the data device
2353 * even though the meta device has failed. For
2354 * now, we behave as though '- -' had been
2355 * set for this device in the table.
2358 dm_put_device(ti
, dev
->data_dev
);
2360 dev
->data_dev
= NULL
;
2363 list_del(&rdev
->same_set
);
2370 if (validate_raid_redundancy(rs
)) {
2371 rs
->ti
->error
= "Insufficient redundancy to activate array";
2376 * Validation of the freshest device provides the source of
2377 * validation for the remaining devices.
2379 rs
->ti
->error
= "Unable to assemble array: Invalid superblocks";
2380 if (super_validate(rs
, freshest
))
2383 rdev_for_each(rdev
, mddev
)
2384 if ((rdev
!= freshest
) && super_validate(rs
, rdev
))
2390 * Adjust data_offset and new_data_offset on all disk members of @rs
2391 * for out of place reshaping if requested by contructor
2393 * We need free space at the beginning of each raid disk for forward
2394 * and at the end for backward reshapes which userspace has to provide
2395 * via remapping/reordering of space.
2397 static int rs_adjust_data_offsets(struct raid_set
*rs
)
2399 sector_t data_offset
= 0, new_data_offset
= 0;
2400 struct md_rdev
*rdev
;
2402 /* Constructor did not request data offset change */
2403 if (!test_bit(__CTR_FLAG_DATA_OFFSET
, &rs
->ctr_flags
)) {
2404 if (!rs_is_reshapable(rs
))
2410 /* HM FIXME: get InSync raid_dev? */
2411 rdev
= &rs
->dev
[0].rdev
;
2413 if (rs
->delta_disks
< 0) {
2415 * Removing disks (reshaping backwards):
2417 * - before reshape: data is at offset 0 and free space
2418 * is at end of each component LV
2420 * - after reshape: data is at offset rs->data_offset != 0 on each component LV
2423 new_data_offset
= rs
->data_offset
;
2425 } else if (rs
->delta_disks
> 0) {
2427 * Adding disks (reshaping forwards):
2429 * - before reshape: data is at offset rs->data_offset != 0 and
2430 * free space is at begin of each component LV
2432 * - after reshape: data is at offset 0 on each component LV
2434 data_offset
= rs
->data_offset
;
2435 new_data_offset
= 0;
2439 * User space passes in 0 for data offset after having removed reshape space
2441 * - or - (data offset != 0)
2443 * Changing RAID layout or chunk size -> toggle offsets
2445 * - before reshape: data is at offset rs->data_offset 0 and
2446 * free space is at end of each component LV
2448 * data is at offset rs->data_offset != 0 and
2449 * free space is at begin of each component LV
2451 * - after reshape: data is at offset 0 if it was at offset != 0
2452 * or at offset != 0 if it was at offset 0
2453 * on each component LV
2456 data_offset
= rs
->data_offset
? rdev
->data_offset
: 0;
2457 new_data_offset
= data_offset
? 0 : rs
->data_offset
;
2458 set_bit(RT_FLAG_UPDATE_SBS
, &rs
->runtime_flags
);
2462 * Make sure we got a minimum amount of free sectors per device
2464 if (rs
->data_offset
&&
2465 to_sector(i_size_read(rdev
->bdev
->bd_inode
)) - rdev
->sectors
< MIN_FREE_RESHAPE_SPACE
) {
2466 rs
->ti
->error
= data_offset
? "No space for forward reshape" :
2467 "No space for backward reshape";
2471 /* Adjust data offsets on all rdevs */
2472 rdev_for_each(rdev
, &rs
->md
) {
2473 rdev
->data_offset
= data_offset
;
2474 rdev
->new_data_offset
= new_data_offset
;
2480 /* Userpace reordered disks -> adjust raid_disk indexes in @rs */
2481 static void __reorder_raid_disk_indexes(struct raid_set
*rs
)
2484 struct md_rdev
*rdev
;
2486 rdev_for_each(rdev
, &rs
->md
) {
2487 rdev
->raid_disk
= i
++;
2488 rdev
->saved_raid_disk
= rdev
->new_raid_disk
= -1;
2493 * Setup @rs for takeover by a different raid level
2495 static int rs_setup_takeover(struct raid_set
*rs
)
2497 struct mddev
*mddev
= &rs
->md
;
2498 struct md_rdev
*rdev
;
2499 unsigned int d
= mddev
->raid_disks
= rs
->raid_disks
;
2500 sector_t new_data_offset
= rs
->dev
[0].rdev
.data_offset
? 0 : rs
->data_offset
;
2502 if (rt_is_raid10(rs
->raid_type
)) {
2503 if (mddev
->level
== 0) {
2504 /* Userpace reordered disks -> adjust raid_disk indexes */
2505 __reorder_raid_disk_indexes(rs
);
2507 /* raid0 -> raid10_far layout */
2508 mddev
->layout
= raid10_format_to_md_layout(rs
, ALGORITHM_RAID10_FAR
,
2510 } else if (mddev
->level
== 1)
2511 /* raid1 -> raid10_near layout */
2512 mddev
->layout
= raid10_format_to_md_layout(rs
, ALGORITHM_RAID10_NEAR
,
2519 clear_bit(MD_ARRAY_FIRST_USE
, &mddev
->flags
);
2520 mddev
->recovery_cp
= MaxSector
;
2523 rdev
= &rs
->dev
[d
].rdev
;
2525 if (test_bit(d
, (void *) rs
->rebuild_disks
)) {
2526 clear_bit(In_sync
, &rdev
->flags
);
2527 clear_bit(Faulty
, &rdev
->flags
);
2528 mddev
->recovery_cp
= rdev
->recovery_offset
= 0;
2529 /* Bitmap has to be created when we do an "up" takeover */
2530 set_bit(MD_ARRAY_FIRST_USE
, &mddev
->flags
);
2533 rdev
->new_data_offset
= new_data_offset
;
2539 /* Prepare @rs for reshape */
2540 static int rs_prepare_reshape(struct raid_set
*rs
)
2543 struct mddev
*mddev
= &rs
->md
;
2545 if (rs_is_raid10(rs
)) {
2546 if (rs
->raid_disks
!= mddev
->raid_disks
&&
2547 __is_raid10_near(mddev
->layout
) &&
2548 rs
->raid10_copies
&&
2549 rs
->raid10_copies
!= __raid10_near_copies(mddev
->layout
)) {
2551 * raid disk have to be multiple of data copies to allow this conversion,
2553 * This is actually not a reshape it is a
2554 * rebuild of any additional mirrors per group
2556 if (rs
->raid_disks
% rs
->raid10_copies
) {
2557 rs
->ti
->error
= "Can't reshape raid10 mirror groups";
2561 /* Userpace reordered disks to add/remove mirrors -> adjust raid_disk indexes */
2562 __reorder_raid_disk_indexes(rs
);
2563 mddev
->layout
= raid10_format_to_md_layout(rs
, ALGORITHM_RAID10_NEAR
,
2565 mddev
->new_layout
= mddev
->layout
;
2570 } else if (rs_is_raid456(rs
))
2573 else if (rs_is_raid1(rs
)) {
2574 if (rs
->delta_disks
) {
2575 /* Process raid1 via delta_disks */
2576 mddev
->degraded
= rs
->delta_disks
< 0 ? -rs
->delta_disks
: rs
->delta_disks
;
2579 /* Process raid1 without delta_disks */
2580 mddev
->raid_disks
= rs
->raid_disks
;
2584 rs
->ti
->error
= "Called with bogus raid type";
2589 set_bit(RT_FLAG_RESHAPE_RS
, &rs
->runtime_flags
);
2590 set_bit(RT_FLAG_UPDATE_SBS
, &rs
->runtime_flags
);
2591 } else if (mddev
->raid_disks
< rs
->raid_disks
)
2592 /* Create new superblocks and bitmaps, if any new disks */
2593 set_bit(RT_FLAG_UPDATE_SBS
, &rs
->runtime_flags
);
2600 * - change raid layout
2601 * - change chunk size
2605 static int rs_setup_reshape(struct raid_set
*rs
)
2608 unsigned int cur_raid_devs
, d
;
2609 struct mddev
*mddev
= &rs
->md
;
2610 struct md_rdev
*rdev
;
2612 mddev
->delta_disks
= rs
->delta_disks
;
2613 cur_raid_devs
= mddev
->raid_disks
;
2615 /* Ignore impossible layout change whilst adding/removing disks */
2616 if (mddev
->delta_disks
&&
2617 mddev
->layout
!= mddev
->new_layout
) {
2618 DMINFO("Ignoring invalid layout change with delta_disks=%d", rs
->delta_disks
);
2619 mddev
->new_layout
= mddev
->layout
;
2623 * Adjust array size:
2625 * - in case of adding disks, array size has
2626 * to grow after the disk adding reshape,
2627 * which'll hapen in the event handler;
2628 * reshape will happen forward, so space has to
2629 * be available at the beginning of each disk
2631 * - in case of removing disks, array size
2632 * has to shrink before starting the reshape,
2633 * which'll happen here;
2634 * reshape will happen backward, so space has to
2635 * be available at the end of each disk
2637 * - data_offset and new_data_offset are
2638 * adjusted for aforementioned out of place
2639 * reshaping based on userspace passing in
2640 * the "data_offset <sectors>" key/value
2641 * pair via the constructor
2645 if (rs
->delta_disks
> 0) {
2646 /* Prepare disks for check in raid4/5/6/10 {check|start}_reshape */
2647 for (d
= cur_raid_devs
; d
< rs
->raid_disks
; d
++) {
2648 rdev
= &rs
->dev
[d
].rdev
;
2649 clear_bit(In_sync
, &rdev
->flags
);
2652 * save_raid_disk needs to be -1, or recovery_offset will be set to 0
2653 * by md, which'll store that erroneously in the superblock on reshape
2655 rdev
->saved_raid_disk
= -1;
2656 rdev
->raid_disk
= d
;
2658 rdev
->sectors
= mddev
->dev_sectors
;
2659 rdev
->recovery_offset
= rs_is_raid1(rs
) ? 0 : MaxSector
;
2662 mddev
->reshape_backwards
= 0; /* adding disks -> forward reshape */
2664 /* Remove disk(s) */
2665 } else if (rs
->delta_disks
< 0) {
2666 r
= rs_set_dev_and_array_sectors(rs
, true);
2667 mddev
->reshape_backwards
= 1; /* removing disk(s) -> backward reshape */
2669 /* Change layout and/or chunk size */
2672 * Reshape layout (e.g. raid5_ls -> raid5_n) and/or chunk size:
2674 * keeping number of disks and do layout change ->
2676 * toggle reshape_backward depending on data_offset:
2678 * - free space upfront -> reshape forward
2680 * - free space at the end -> reshape backward
2683 * This utilizes free reshape space avoiding the need
2684 * for userspace to move (parts of) LV segments in
2685 * case of layout/chunksize change (for disk
2686 * adding/removing reshape space has to be at
2687 * the proper address (see above with delta_disks):
2689 * add disk(s) -> begin
2690 * remove disk(s)-> end
2692 mddev
->reshape_backwards
= rs
->dev
[0].rdev
.data_offset
? 0 : 1;
2699 * Enable/disable discard support on RAID set depending on
2700 * RAID level and discard properties of underlying RAID members.
2702 static void configure_discard_support(struct raid_set
*rs
)
2706 struct dm_target
*ti
= rs
->ti
;
2708 /* Assume discards not supported until after checks below. */
2709 ti
->discards_supported
= false;
2711 /* RAID level 4,5,6 require discard_zeroes_data for data integrity! */
2712 raid456
= (rs
->md
.level
== 4 || rs
->md
.level
== 5 || rs
->md
.level
== 6);
2714 for (i
= 0; i
< rs
->raid_disks
; i
++) {
2715 struct request_queue
*q
;
2717 if (!rs
->dev
[i
].rdev
.bdev
)
2720 q
= bdev_get_queue(rs
->dev
[i
].rdev
.bdev
);
2721 if (!q
|| !blk_queue_discard(q
))
2725 if (!q
->limits
.discard_zeroes_data
)
2727 if (!devices_handle_discard_safely
) {
2728 DMERR("raid456 discard support disabled due to discard_zeroes_data uncertainty.");
2729 DMERR("Set dm-raid.devices_handle_discard_safely=Y to override.");
2735 /* All RAID members properly support discards */
2736 ti
->discards_supported
= true;
2739 * RAID1 and RAID10 personalities require bio splitting,
2740 * RAID0/4/5/6 don't and process large discard bios properly.
2742 ti
->split_discard_bios
= !!(rs
->md
.level
== 1 || rs
->md
.level
== 10);
2743 ti
->num_discard_bios
= 1;
2747 * Construct a RAID0/1/10/4/5/6 mapping:
2749 * <raid_type> <#raid_params> <raid_params>{0,} \
2750 * <#raid_devs> [<meta_dev1> <dev1>]{1,}
2752 * <raid_params> varies by <raid_type>. See 'parse_raid_params' for
2753 * details on possible <raid_params>.
2755 * Userspace is free to initialize the metadata devices, hence the superblocks to
2756 * enforce recreation based on the passed in table parameters.
2759 static int raid_ctr(struct dm_target
*ti
, unsigned int argc
, char **argv
)
2763 struct raid_type
*rt
;
2764 unsigned int num_raid_params
, num_raid_devs
;
2765 sector_t calculated_dev_sectors
;
2766 struct raid_set
*rs
= NULL
;
2768 struct rs_layout rs_layout
;
2769 struct dm_arg_set as
= { argc
, argv
}, as_nrd
;
2770 struct dm_arg _args
[] = {
2771 { 0, as
.argc
, "Cannot understand number of raid parameters" },
2772 { 1, 254, "Cannot understand number of raid devices parameters" }
2775 /* Must have <raid_type> */
2776 arg
= dm_shift_arg(&as
);
2778 ti
->error
= "No arguments";
2782 rt
= get_raid_type(arg
);
2784 ti
->error
= "Unrecognised raid_type";
2788 /* Must have <#raid_params> */
2789 if (dm_read_arg_group(_args
, &as
, &num_raid_params
, &ti
->error
))
2792 /* number of raid device tupples <meta_dev data_dev> */
2794 dm_consume_args(&as_nrd
, num_raid_params
);
2795 _args
[1].max
= (as_nrd
.argc
- 1) / 2;
2796 if (dm_read_arg(_args
+ 1, &as_nrd
, &num_raid_devs
, &ti
->error
))
2799 if (!__within_range(num_raid_devs
, 1, MAX_RAID_DEVICES
)) {
2800 ti
->error
= "Invalid number of supplied raid devices";
2804 rs
= raid_set_alloc(ti
, rt
, num_raid_devs
);
2808 r
= parse_raid_params(rs
, &as
, num_raid_params
);
2812 r
= parse_dev_params(rs
, &as
);
2816 rs
->md
.sync_super
= super_sync
;
2819 * Calculate ctr requested array and device sizes to allow
2820 * for superblock analysis needing device sizes defined.
2822 * Any existing superblock will overwrite the array and device sizes
2824 r
= rs_set_dev_and_array_sectors(rs
, false);
2828 calculated_dev_sectors
= rs
->dev
[0].rdev
.sectors
;
2831 * Backup any new raid set level, layout, ...
2832 * requested to be able to compare to superblock
2833 * members for conversion decisions.
2835 rs_config_backup(rs
, &rs_layout
);
2837 r
= analyse_superblocks(ti
, rs
);
2841 resize
= calculated_dev_sectors
!= rs
->dev
[0].rdev
.sectors
;
2843 INIT_WORK(&rs
->md
.event_work
, do_table_event
);
2845 ti
->num_flush_bios
= 1;
2847 /* Restore any requested new layout for conversion decision */
2848 rs_config_restore(rs
, &rs_layout
);
2851 * Now that we have any superblock metadata available,
2852 * check for new, recovering, reshaping, to be taken over,
2853 * to be reshaped or an existing, unchanged raid set to
2856 if (test_bit(MD_ARRAY_FIRST_USE
, &rs
->md
.flags
)) {
2857 /* A new raid6 set has to be recovered to ensure proper parity and Q-Syndrome */
2858 if (rs_is_raid6(rs
) &&
2859 test_bit(__CTR_FLAG_NOSYNC
, &rs
->ctr_flags
)) {
2860 ti
->error
= "'nosync' not allowed for new raid6 set";
2864 rs_setup_recovery(rs
, 0);
2865 set_bit(RT_FLAG_UPDATE_SBS
, &rs
->runtime_flags
);
2867 } else if (rs_is_recovering(rs
)) {
2868 /* A recovering raid set may be resized */
2869 ; /* skip setup rs */
2870 } else if (rs_is_reshaping(rs
)) {
2871 /* Have to reject size change request during reshape */
2873 ti
->error
= "Can't resize a reshaping raid set";
2878 } else if (rs_takeover_requested(rs
)) {
2879 if (rs_is_reshaping(rs
)) {
2880 ti
->error
= "Can't takeover a reshaping raid set";
2886 * If a takeover is needed, userspace sets any additional
2887 * devices to rebuild and we can check for a valid request here.
2889 * If acceptible, set the level to the new requested
2890 * one, prohibit requesting recovery, allow the raid
2891 * set to run and store superblocks during resume.
2893 r
= rs_check_takeover(rs
);
2897 r
= rs_setup_takeover(rs
);
2901 set_bit(RT_FLAG_UPDATE_SBS
, &rs
->runtime_flags
);
2902 /* Takeover ain't recovery, so disable recovery */
2903 rs_setup_recovery(rs
, MaxSector
);
2905 } else if (rs_reshape_requested(rs
)) {
2907 * We can only prepare for a reshape here, because the
2908 * raid set needs to run to provide the repective reshape
2909 * check functions via its MD personality instance.
2911 * So do the reshape check after md_run() succeeded.
2913 r
= rs_prepare_reshape(rs
);
2917 /* Reshaping ain't recovery, so disable recovery */
2918 rs_setup_recovery(rs
, MaxSector
);
2921 /* May not set recovery when a device rebuild is requested */
2922 if (test_bit(__CTR_FLAG_REBUILD
, &rs
->ctr_flags
)) {
2923 rs_setup_recovery(rs
, MaxSector
);
2924 set_bit(RT_FLAG_UPDATE_SBS
, &rs
->runtime_flags
);
2926 rs_setup_recovery(rs
, test_bit(__CTR_FLAG_SYNC
, &rs
->ctr_flags
) ?
2927 0 : (resize
? calculated_dev_sectors
: MaxSector
));
2931 /* If constructor requested it, change data and new_data offsets */
2932 r
= rs_adjust_data_offsets(rs
);
2936 /* Start raid set read-only and assumed clean to change in raid_resume() */
2939 set_bit(MD_RECOVERY_FROZEN
, &rs
->md
.recovery
);
2941 /* Has to be held on running the array */
2942 mddev_lock_nointr(&rs
->md
);
2943 r
= md_run(&rs
->md
);
2944 rs
->md
.in_sync
= 0; /* Assume already marked dirty */
2947 ti
->error
= "Failed to run raid array";
2948 mddev_unlock(&rs
->md
);
2952 rs
->callbacks
.congested_fn
= raid_is_congested
;
2953 dm_table_add_target_callbacks(ti
->table
, &rs
->callbacks
);
2955 mddev_suspend(&rs
->md
);
2957 /* Try to adjust the raid4/5/6 stripe cache size to the stripe size */
2958 if (rs_is_raid456(rs
)) {
2959 r
= rs_set_raid456_stripe_cache(rs
);
2961 goto bad_stripe_cache
;
2964 /* Now do an early reshape check */
2965 if (test_bit(RT_FLAG_RESHAPE_RS
, &rs
->runtime_flags
)) {
2966 r
= rs_check_reshape(rs
);
2968 goto bad_check_reshape
;
2970 /* Restore new, ctr requested layout to perform check */
2971 rs_config_restore(rs
, &rs_layout
);
2973 if (rs
->md
.pers
->start_reshape
) {
2974 r
= rs
->md
.pers
->check_reshape(&rs
->md
);
2976 ti
->error
= "Reshape check failed";
2977 goto bad_check_reshape
;
2982 mddev_unlock(&rs
->md
);
2994 static void raid_dtr(struct dm_target
*ti
)
2996 struct raid_set
*rs
= ti
->private;
2998 list_del_init(&rs
->callbacks
.list
);
3003 static int raid_map(struct dm_target
*ti
, struct bio
*bio
)
3005 struct raid_set
*rs
= ti
->private;
3006 struct mddev
*mddev
= &rs
->md
;
3009 * If we're reshaping to add disk(s)), ti->len and
3010 * mddev->array_sectors will differ during the process
3011 * (ti->len > mddev->array_sectors), so we have to requeue
3012 * bios with addresses > mddev->array_sectors here or
3013 * there will occur accesses past EOD of the component
3014 * data images thus erroring the raid set.
3016 if (unlikely(bio_end_sector(bio
) > mddev
->array_sectors
))
3017 return DM_MAPIO_REQUEUE
;
3019 mddev
->pers
->make_request(mddev
, bio
);
3021 return DM_MAPIO_SUBMITTED
;
3024 /* Return string describing the current sync action of @mddev */
3025 static const char *decipher_sync_action(struct mddev
*mddev
)
3027 if (test_bit(MD_RECOVERY_FROZEN
, &mddev
->recovery
))
3030 if (test_bit(MD_RECOVERY_RUNNING
, &mddev
->recovery
) ||
3031 (!mddev
->ro
&& test_bit(MD_RECOVERY_NEEDED
, &mddev
->recovery
))) {
3032 if (test_bit(MD_RECOVERY_RESHAPE
, &mddev
->recovery
))
3035 if (test_bit(MD_RECOVERY_SYNC
, &mddev
->recovery
)) {
3036 if (!test_bit(MD_RECOVERY_REQUESTED
, &mddev
->recovery
))
3038 else if (test_bit(MD_RECOVERY_CHECK
, &mddev
->recovery
))
3043 if (test_bit(MD_RECOVERY_RECOVER
, &mddev
->recovery
))
3051 * Return status string @rdev
3053 * Status characters:
3055 * 'D' = Dead/Failed device
3056 * 'a' = Alive but not in-sync
3057 * 'A' = Alive and in-sync
3059 static const char *__raid_dev_status(struct md_rdev
*rdev
, bool array_in_sync
)
3061 if (test_bit(Faulty
, &rdev
->flags
))
3063 else if (!array_in_sync
|| !test_bit(In_sync
, &rdev
->flags
))
3069 /* Helper to return resync/reshape progress for @rs and @array_in_sync */
3070 static sector_t
rs_get_progress(struct raid_set
*rs
,
3071 sector_t resync_max_sectors
, bool *array_in_sync
)
3073 sector_t r
, recovery_cp
, curr_resync_completed
;
3074 struct mddev
*mddev
= &rs
->md
;
3076 curr_resync_completed
= mddev
->curr_resync_completed
?: mddev
->recovery_cp
;
3077 recovery_cp
= mddev
->recovery_cp
;
3078 *array_in_sync
= false;
3080 if (rs_is_raid0(rs
)) {
3081 r
= resync_max_sectors
;
3082 *array_in_sync
= true;
3085 r
= mddev
->reshape_position
;
3087 /* Reshape is relative to the array size */
3088 if (test_bit(MD_RECOVERY_RESHAPE
, &mddev
->recovery
) ||
3090 if (r
== MaxSector
) {
3091 *array_in_sync
= true;
3092 r
= resync_max_sectors
;
3094 /* Got to reverse on backward reshape */
3095 if (mddev
->reshape_backwards
)
3096 r
= mddev
->array_sectors
- r
;
3098 /* Devide by # of data stripes */
3099 sector_div(r
, mddev_data_stripes(rs
));
3102 /* Sync is relative to the component device size */
3103 } else if (test_bit(MD_RECOVERY_RUNNING
, &mddev
->recovery
))
3104 r
= curr_resync_completed
;
3108 if (r
== MaxSector
) {
3112 *array_in_sync
= true;
3113 r
= resync_max_sectors
;
3114 } else if (test_bit(MD_RECOVERY_REQUESTED
, &mddev
->recovery
)) {
3116 * If "check" or "repair" is occurring, the raid set has
3117 * undergone an initial sync and the health characters
3118 * should not be 'a' anymore.
3120 *array_in_sync
= true;
3122 struct md_rdev
*rdev
;
3125 * The raid set may be doing an initial sync, or it may
3126 * be rebuilding individual components. If all the
3127 * devices are In_sync, then it is the raid set that is
3128 * being initialized.
3130 rdev_for_each(rdev
, mddev
)
3131 if (!test_bit(In_sync
, &rdev
->flags
))
3132 *array_in_sync
= true;
3134 r
= 0; /* HM FIXME: TESTME: https://bugzilla.redhat.com/show_bug.cgi?id=1210637 ? */
3142 /* Helper to return @dev name or "-" if !@dev */
3143 static const char *__get_dev_name(struct dm_dev
*dev
)
3145 return dev
? dev
->name
: "-";
3148 static void raid_status(struct dm_target
*ti
, status_type_t type
,
3149 unsigned int status_flags
, char *result
, unsigned int maxlen
)
3151 struct raid_set
*rs
= ti
->private;
3152 struct mddev
*mddev
= &rs
->md
;
3153 struct r5conf
*conf
= mddev
->private;
3154 int i
, max_nr_stripes
= conf
? conf
->max_nr_stripes
: 0;
3156 unsigned int raid_param_cnt
= 1; /* at least 1 for chunksize */
3157 unsigned int sz
= 0;
3158 unsigned int rebuild_disks
;
3159 unsigned int write_mostly_params
= 0;
3160 sector_t progress
, resync_max_sectors
, resync_mismatches
;
3161 const char *sync_action
;
3162 struct raid_type
*rt
;
3163 struct md_rdev
*rdev
;
3166 case STATUSTYPE_INFO
:
3167 /* *Should* always succeed */
3168 rt
= get_raid_type_by_ll(mddev
->new_level
, mddev
->new_layout
);
3172 DMEMIT("%s %d ", rt
->name
, mddev
->raid_disks
);
3174 /* Access most recent mddev properties for status output */
3176 /* Get sensible max sectors even if raid set not yet started */
3177 resync_max_sectors
= test_bit(RT_FLAG_RS_PRERESUMED
, &rs
->runtime_flags
) ?
3178 mddev
->resync_max_sectors
: mddev
->dev_sectors
;
3179 progress
= rs_get_progress(rs
, resync_max_sectors
, &array_in_sync
);
3180 resync_mismatches
= (mddev
->last_sync_action
&& !strcasecmp(mddev
->last_sync_action
, "check")) ?
3181 atomic64_read(&mddev
->resync_mismatches
) : 0;
3182 sync_action
= decipher_sync_action(&rs
->md
);
3184 /* HM FIXME: do we want another state char for raid0? It shows 'D' or 'A' now */
3185 rdev_for_each(rdev
, mddev
)
3186 DMEMIT(__raid_dev_status(rdev
, array_in_sync
));
3189 * In-sync/Reshape ratio:
3190 * The in-sync ratio shows the progress of:
3191 * - Initializing the raid set
3192 * - Rebuilding a subset of devices of the raid set
3193 * The user can distinguish between the two by referring
3194 * to the status characters.
3196 * The reshape ratio shows the progress of
3197 * changing the raid layout or the number of
3198 * disks of a raid set
3200 DMEMIT(" %llu/%llu", (unsigned long long) progress
,
3201 (unsigned long long) resync_max_sectors
);
3207 * See Documentation/device-mapper/dm-raid.txt for
3208 * information on each of these states.
3210 DMEMIT(" %s", sync_action
);
3215 * resync_mismatches/mismatch_cnt
3216 * This field shows the number of discrepancies found when
3217 * performing a "check" of the raid set.
3219 DMEMIT(" %llu", (unsigned long long) resync_mismatches
);
3224 * data_offset (needed for out of space reshaping)
3225 * This field shows the data offset into the data
3226 * image LV where the first stripes data starts.
3228 * We keep data_offset equal on all raid disks of the set,
3229 * so retrieving it from the first raid disk is sufficient.
3231 DMEMIT(" %llu", (unsigned long long) rs
->dev
[0].rdev
.data_offset
);
3234 case STATUSTYPE_TABLE
:
3235 /* Report the table line string you would use to construct this raid set */
3237 /* Calculate raid parameter count */
3238 for (i
= 0; i
< rs
->raid_disks
; i
++)
3239 if (test_bit(WriteMostly
, &rs
->dev
[i
].rdev
.flags
))
3240 write_mostly_params
+= 2;
3241 rebuild_disks
= memweight(rs
->rebuild_disks
, DISKS_ARRAY_ELEMS
* sizeof(*rs
->rebuild_disks
));
3242 raid_param_cnt
+= rebuild_disks
* 2 +
3243 write_mostly_params
+
3244 hweight32(rs
->ctr_flags
& CTR_FLAG_OPTIONS_NO_ARGS
) +
3245 hweight32(rs
->ctr_flags
& CTR_FLAG_OPTIONS_ONE_ARG
) * 2;
3246 /* Emit table line */
3247 DMEMIT("%s %u %u", rs
->raid_type
->name
, raid_param_cnt
, mddev
->new_chunk_sectors
);
3248 if (test_bit(__CTR_FLAG_RAID10_FORMAT
, &rs
->ctr_flags
))
3249 DMEMIT(" %s %s", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT
),
3250 raid10_md_layout_to_format(mddev
->layout
));
3251 if (test_bit(__CTR_FLAG_RAID10_COPIES
, &rs
->ctr_flags
))
3252 DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES
),
3253 raid10_md_layout_to_copies(mddev
->layout
));
3254 if (test_bit(__CTR_FLAG_NOSYNC
, &rs
->ctr_flags
))
3255 DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC
));
3256 if (test_bit(__CTR_FLAG_SYNC
, &rs
->ctr_flags
))
3257 DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_SYNC
));
3258 if (test_bit(__CTR_FLAG_REGION_SIZE
, &rs
->ctr_flags
))
3259 DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE
),
3260 (unsigned long long) to_sector(mddev
->bitmap_info
.chunksize
));
3261 if (test_bit(__CTR_FLAG_DATA_OFFSET
, &rs
->ctr_flags
))
3262 DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET
),
3263 (unsigned long long) rs
->data_offset
);
3264 if (test_bit(__CTR_FLAG_DAEMON_SLEEP
, &rs
->ctr_flags
))
3265 DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP
),
3266 mddev
->bitmap_info
.daemon_sleep
);
3267 if (test_bit(__CTR_FLAG_DELTA_DISKS
, &rs
->ctr_flags
))
3268 DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS
),
3269 max(rs
->delta_disks
, mddev
->delta_disks
));
3270 if (test_bit(__CTR_FLAG_STRIPE_CACHE
, &rs
->ctr_flags
))
3271 DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE
),
3274 for (i
= 0; i
< rs
->raid_disks
; i
++)
3275 if (test_bit(rs
->dev
[i
].rdev
.raid_disk
, (void *) rs
->rebuild_disks
))
3276 DMEMIT(" %s %u", dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD
),
3277 rs
->dev
[i
].rdev
.raid_disk
);
3278 if (write_mostly_params
)
3279 for (i
= 0; i
< rs
->raid_disks
; i
++)
3280 if (test_bit(WriteMostly
, &rs
->dev
[i
].rdev
.flags
))
3281 DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY
),
3282 rs
->dev
[i
].rdev
.raid_disk
);
3283 if (test_bit(__CTR_FLAG_MAX_WRITE_BEHIND
, &rs
->ctr_flags
))
3284 DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND
),
3285 mddev
->bitmap_info
.max_write_behind
);
3286 if (test_bit(__CTR_FLAG_MAX_RECOVERY_RATE
, &rs
->ctr_flags
))
3287 DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE
),
3288 mddev
->sync_speed_max
);
3289 if (test_bit(__CTR_FLAG_MIN_RECOVERY_RATE
, &rs
->ctr_flags
))
3290 DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE
),
3291 mddev
->sync_speed_min
);
3292 DMEMIT(" %d", rs
->raid_disks
);
3293 for (i
= 0; i
< rs
->raid_disks
; i
++)
3294 DMEMIT(" %s %s", __get_dev_name(rs
->dev
[i
].meta_dev
),
3295 __get_dev_name(rs
->dev
[i
].data_dev
));
3299 static int raid_message(struct dm_target
*ti
, unsigned int argc
, char **argv
)
3301 struct raid_set
*rs
= ti
->private;
3302 struct mddev
*mddev
= &rs
->md
;
3304 if (!mddev
->pers
|| !mddev
->pers
->sync_request
)
3307 if (!strcasecmp(argv
[0], "frozen"))
3308 set_bit(MD_RECOVERY_FROZEN
, &mddev
->recovery
);
3310 clear_bit(MD_RECOVERY_FROZEN
, &mddev
->recovery
);
3312 if (!strcasecmp(argv
[0], "idle") || !strcasecmp(argv
[0], "frozen")) {
3313 if (mddev
->sync_thread
) {
3314 set_bit(MD_RECOVERY_INTR
, &mddev
->recovery
);
3315 md_reap_sync_thread(mddev
);
3317 } else if (test_bit(MD_RECOVERY_RUNNING
, &mddev
->recovery
) ||
3318 test_bit(MD_RECOVERY_NEEDED
, &mddev
->recovery
))
3320 else if (!strcasecmp(argv
[0], "resync"))
3321 ; /* MD_RECOVERY_NEEDED set below */
3322 else if (!strcasecmp(argv
[0], "recover"))
3323 set_bit(MD_RECOVERY_RECOVER
, &mddev
->recovery
);
3325 if (!strcasecmp(argv
[0], "check"))
3326 set_bit(MD_RECOVERY_CHECK
, &mddev
->recovery
);
3327 else if (!!strcasecmp(argv
[0], "repair"))
3329 set_bit(MD_RECOVERY_REQUESTED
, &mddev
->recovery
);
3330 set_bit(MD_RECOVERY_SYNC
, &mddev
->recovery
);
3332 if (mddev
->ro
== 2) {
3333 /* A write to sync_action is enough to justify
3334 * canceling read-auto mode
3337 if (!mddev
->suspended
&& mddev
->sync_thread
)
3338 md_wakeup_thread(mddev
->sync_thread
);
3340 set_bit(MD_RECOVERY_NEEDED
, &mddev
->recovery
);
3341 if (!mddev
->suspended
&& mddev
->thread
)
3342 md_wakeup_thread(mddev
->thread
);
3347 static int raid_iterate_devices(struct dm_target
*ti
,
3348 iterate_devices_callout_fn fn
, void *data
)
3350 struct raid_set
*rs
= ti
->private;
3354 for (i
= 0; !r
&& i
< rs
->md
.raid_disks
; i
++)
3355 if (rs
->dev
[i
].data_dev
)
3357 rs
->dev
[i
].data_dev
,
3358 0, /* No offset on data devs */
3365 static void raid_io_hints(struct dm_target
*ti
, struct queue_limits
*limits
)
3367 struct raid_set
*rs
= ti
->private;
3368 unsigned int chunk_size
= to_bytes(rs
->md
.chunk_sectors
);
3370 blk_limits_io_min(limits
, chunk_size
);
3371 blk_limits_io_opt(limits
, chunk_size
* mddev_data_stripes(rs
));
3374 static void raid_presuspend(struct dm_target
*ti
)
3376 struct raid_set
*rs
= ti
->private;
3378 md_stop_writes(&rs
->md
);
3381 static void raid_postsuspend(struct dm_target
*ti
)
3383 struct raid_set
*rs
= ti
->private;
3385 if (!rs
->md
.suspended
)
3386 mddev_suspend(&rs
->md
);
3391 static void attempt_restore_of_faulty_devices(struct raid_set
*rs
)
3394 uint64_t failed_devices
, cleared_failed_devices
= 0;
3395 unsigned long flags
;
3396 struct dm_raid_superblock
*sb
;
3399 for (i
= 0; i
< rs
->md
.raid_disks
; i
++) {
3400 r
= &rs
->dev
[i
].rdev
;
3401 if (test_bit(Faulty
, &r
->flags
) && r
->sb_page
&&
3402 sync_page_io(r
, 0, r
->sb_size
, r
->sb_page
,
3403 REQ_OP_READ
, 0, true)) {
3404 DMINFO("Faulty %s device #%d has readable super block."
3405 " Attempting to revive it.",
3406 rs
->raid_type
->name
, i
);
3409 * Faulty bit may be set, but sometimes the array can
3410 * be suspended before the personalities can respond
3411 * by removing the device from the array (i.e. calling
3412 * 'hot_remove_disk'). If they haven't yet removed
3413 * the failed device, its 'raid_disk' number will be
3414 * '>= 0' - meaning we must call this function
3417 if ((r
->raid_disk
>= 0) &&
3418 (r
->mddev
->pers
->hot_remove_disk(r
->mddev
, r
) != 0))
3419 /* Failed to revive this device, try next */
3423 r
->saved_raid_disk
= i
;
3425 clear_bit(Faulty
, &r
->flags
);
3426 clear_bit(WriteErrorSeen
, &r
->flags
);
3427 clear_bit(In_sync
, &r
->flags
);
3428 if (r
->mddev
->pers
->hot_add_disk(r
->mddev
, r
)) {
3430 r
->saved_raid_disk
= -1;
3433 r
->recovery_offset
= 0;
3434 cleared_failed_devices
|= 1 << i
;
3438 if (cleared_failed_devices
) {
3439 rdev_for_each(r
, &rs
->md
) {
3440 sb
= page_address(r
->sb_page
);
3441 failed_devices
= le64_to_cpu(sb
->failed_devices
);
3442 failed_devices
&= ~cleared_failed_devices
;
3443 sb
->failed_devices
= cpu_to_le64(failed_devices
);
3448 static int __load_dirty_region_bitmap(struct raid_set
*rs
)
3452 /* Try loading the bitmap unless "raid0", which does not have one */
3453 if (!rs_is_raid0(rs
) &&
3454 !test_and_set_bit(RT_FLAG_RS_BITMAP_LOADED
, &rs
->runtime_flags
)) {
3455 r
= bitmap_load(&rs
->md
);
3457 DMERR("Failed to load bitmap");
3463 /* Enforce updating all superblocks */
3464 static void rs_update_sbs(struct raid_set
*rs
)
3466 struct mddev
*mddev
= &rs
->md
;
3469 set_bit(MD_CHANGE_DEVS
, &mddev
->flags
);
3471 md_update_sb(mddev
, 1);
3476 * Reshape changes raid algorithm of @rs to new one within personality
3477 * (e.g. raid6_zr -> raid6_nc), changes stripe size, adds/removes
3478 * disks from a raid set thus growing/shrinking it or resizes the set
3480 * Call mddev_lock_nointr() before!
3482 static int rs_start_reshape(struct raid_set
*rs
)
3485 struct mddev
*mddev
= &rs
->md
;
3486 struct md_personality
*pers
= mddev
->pers
;
3488 r
= rs_setup_reshape(rs
);
3492 /* Need to be resumed to be able to start reshape, recovery is frozen until raid_resume() though */
3493 if (mddev
->suspended
)
3494 mddev_resume(mddev
);
3497 * Check any reshape constraints enforced by the personalility
3499 * May as well already kick the reshape off so that * pers->start_reshape() becomes optional.
3501 r
= pers
->check_reshape(mddev
);
3503 rs
->ti
->error
= "pers->check_reshape() failed";
3508 * Personality may not provide start reshape method in which
3509 * case check_reshape above has already covered everything
3511 if (pers
->start_reshape
) {
3512 r
= pers
->start_reshape(mddev
);
3514 rs
->ti
->error
= "pers->start_reshape() failed";
3519 /* Suspend because a resume will happen in raid_resume() */
3520 if (!mddev
->suspended
)
3521 mddev_suspend(mddev
);
3524 * Now reshape got set up, update superblocks to
3525 * reflect the fact so that a table reload will
3526 * access proper superblock content in the ctr.
3533 static int raid_preresume(struct dm_target
*ti
)
3536 struct raid_set
*rs
= ti
->private;
3537 struct mddev
*mddev
= &rs
->md
;
3539 /* This is a resume after a suspend of the set -> it's already started */
3540 if (test_and_set_bit(RT_FLAG_RS_PRERESUMED
, &rs
->runtime_flags
))
3544 * The superblocks need to be updated on disk if the
3545 * array is new or new devices got added (thus zeroed
3546 * out by userspace) or __load_dirty_region_bitmap
3547 * will overwrite them in core with old data or fail.
3549 if (test_bit(RT_FLAG_UPDATE_SBS
, &rs
->runtime_flags
))
3553 * Disable/enable discard support on raid set after any
3554 * conversion, because devices can have been added
3556 configure_discard_support(rs
);
3558 /* Load the bitmap from disk unless raid0 */
3559 r
= __load_dirty_region_bitmap(rs
);
3563 /* Resize bitmap to adjust to changed region size (aka MD bitmap chunksize) */
3564 if (test_bit(RT_FLAG_RS_BITMAP_LOADED
, &rs
->runtime_flags
) &&
3565 mddev
->bitmap_info
.chunksize
!= to_bytes(rs
->requested_bitmap_chunk_sectors
)) {
3566 r
= bitmap_resize(mddev
->bitmap
, mddev
->dev_sectors
,
3567 to_bytes(rs
->requested_bitmap_chunk_sectors
), 0);
3569 DMERR("Failed to resize bitmap");
3572 /* Check for any resize/reshape on @rs and adjust/initiate */
3573 /* Be prepared for mddev_resume() in raid_resume() */
3574 set_bit(MD_RECOVERY_FROZEN
, &mddev
->recovery
);
3575 if (mddev
->recovery_cp
&& mddev
->recovery_cp
< MaxSector
) {
3576 set_bit(MD_RECOVERY_SYNC
, &mddev
->recovery
);
3577 mddev
->resync_min
= mddev
->recovery_cp
;
3580 rs_set_capacity(rs
);
3582 /* Check for any reshape request unless new raid set */
3583 if (test_and_clear_bit(RT_FLAG_RESHAPE_RS
, &rs
->runtime_flags
)) {
3584 /* Initiate a reshape. */
3585 mddev_lock_nointr(mddev
);
3586 r
= rs_start_reshape(rs
);
3587 mddev_unlock(mddev
);
3589 DMWARN("Failed to check/start reshape, continuing without change");
3596 static void raid_resume(struct dm_target
*ti
)
3598 struct raid_set
*rs
= ti
->private;
3599 struct mddev
*mddev
= &rs
->md
;
3601 if (test_and_set_bit(RT_FLAG_RS_RESUMED
, &rs
->runtime_flags
)) {
3603 * A secondary resume while the device is active.
3604 * Take this opportunity to check whether any failed
3605 * devices are reachable again.
3607 attempt_restore_of_faulty_devices(rs
);
3613 clear_bit(MD_RECOVERY_FROZEN
, &mddev
->recovery
);
3615 if (mddev
->suspended
)
3616 mddev_resume(mddev
);
3619 static struct target_type raid_target
= {
3621 .version
= {1, 9, 0},
3622 .module
= THIS_MODULE
,
3626 .status
= raid_status
,
3627 .message
= raid_message
,
3628 .iterate_devices
= raid_iterate_devices
,
3629 .io_hints
= raid_io_hints
,
3630 .presuspend
= raid_presuspend
,
3631 .postsuspend
= raid_postsuspend
,
3632 .preresume
= raid_preresume
,
3633 .resume
= raid_resume
,
3636 static int __init
dm_raid_init(void)
3638 DMINFO("Loading target version %u.%u.%u",
3639 raid_target
.version
[0],
3640 raid_target
.version
[1],
3641 raid_target
.version
[2]);
3642 return dm_register_target(&raid_target
);
3645 static void __exit
dm_raid_exit(void)
3647 dm_unregister_target(&raid_target
);
3650 module_init(dm_raid_init
);
3651 module_exit(dm_raid_exit
);
3653 module_param(devices_handle_discard_safely
, bool, 0644);
3654 MODULE_PARM_DESC(devices_handle_discard_safely
,
3655 "Set to Y if all devices in each array reliably return zeroes on reads from discarded regions");
3657 MODULE_DESCRIPTION(DM_NAME
" raid0/1/10/4/5/6 target");
3658 MODULE_ALIAS("dm-raid0");
3659 MODULE_ALIAS("dm-raid1");
3660 MODULE_ALIAS("dm-raid10");
3661 MODULE_ALIAS("dm-raid4");
3662 MODULE_ALIAS("dm-raid5");
3663 MODULE_ALIAS("dm-raid6");
3664 MODULE_AUTHOR("Neil Brown <dm-devel@redhat.com>");
3665 MODULE_AUTHOR("Heinz Mauelshagen <dm-devel@redhat.com>");
3666 MODULE_LICENSE("GPL");